1 s2.0-s0045790610000182-main

Computers and Electrical Engineering 36 (2010) 978–992

Contents lists available at ScienceDirect

Computers and Electrical Engineering

journal homepage: www.elsevier .com/ locate /compeleceng

A CDMA/TDD approach for wireless mesh networks q

Imam Al-wazedi *, Ahmed K. ElhakeemDepartment of ECE, Concordia University, Montreal, Quebec, Canada

a r t i c l e i n f o

Article history:Received 1 April 2009Received in revised form 31 October 2009Accepted 5 February 2010Available online 29 March 2010

Keywords:WCDMATDMATDDMesh networks

0045-7906/$ - see front matter � 2010 Elsevier Ltddoi:10.1016/j.compeleceng.2010.02.003

q Reviews processed and proposed for publication* Corresponding author.

E-mail addresses: [email protected] (I.

a b s t r a c t

Wide band mesh or star oriented networks have recently become a subject of greater inter-est. Providing wideband multimedia access for a variety of applications has led to theinception of mesh networks. Classic access techniques such as FDMA and TDMA have beenthe norm for such networks. CDMA maximum transmitter power is much less than TDMAand FDMA counter parts, which is an important asset for mobile operation. In this paper weintroduce a code division multiple access/time division duplex technique CDMA/TDD forsuch networks. The CDMA approach is an almost play and plug technology for wirelessaccess, making it amenable for implementation by the mesh network service station, SS.Further it inherently allows mesh network service stations to use a combination of turbocoding and dynamic parallel orthogonal transmission to improve network efficiency. Weoutline briefly the new transmitter and receiver structures then evaluate the efficiency,delay and delay jitter. By analysis we show the advantages over classic counter parts withrespect to the total network efficiency achievable especially for larger number of hops.

� 2010 Elsevier Ltd. All rights reserved.

1. Introduction

The performance of cross layer approaches in wireless mesh networks such as WIMAX has been subject of intensive re-search [3,4]. The general PHY, MAC layers and views on several cross layer issues related to WIMAX, especially the OFDMA/TDD system are given in [1]. In wireless metropolitan area network (MAN), OFDMA PHY layer is based on OFDM modulation.TDD systems use the same frequency band for downlink and uplink and the frames are divided into the DL sub-frames andUL sub-frames in the time domain. Mesh networks such as WIMAX is capable of using both FDD and TDD connections. Toprovide highest transport efficiency in broad band networks, time division duplex (TDD) is preferred over FDD because itoffers more flexibility in changing the UL an DL bandwidth ratio according to the dynamic traffic pattern [5]. The MAC pro-tocol of IEEE 802.16 is connection oriented and each connection is identified by connection identification number (CID),which is given to each SS (subscriber stations) in the initialization process. In this protocol the SS use TDMA on the uplinkand transmit back to BS in a specific time slot [2].

In IEEE 802.16 Mesh mode, Mesh Network Configuration (MSH-NCFG) and Mesh Network Entry (MSH-NENT) messagesare used for advertisement of the mesh network and for helping new nodes to synchronize and to joining the mesh network.Active nodes within the mesh periodically advertise MSH-NCFG messages with Network Descriptor, which outlines the basicnetwork configuration information such as BS ID number and the base channel currently used. A new node that plans to joinan active mesh network scans for active networks and listens to MSH-NCFG message. The new node establishes coarse syn-chronization and starts the network entry process based on the information given by MSH-NCFG. Among all possible neigh-bors that advertise MSH-NCFG, the joining node (which is called Candidate Node in the 802.16 Mesh mode terminologies)

. All rights reserved.

to the Editor-in-Chief by Associate Editor Dr. A. El-Osery.

Al-wazedi), [email protected] (A.K. Elhakeem).

http://dx.doi.org/10.1016/j.compeleceng.2010.02.003

mailto:[email protected]

mailto:[email protected]

http://www.sciencedirect.com/science/journal/00457906

http://www.elsevier.com/locate/compeleceng

I. Al-wazedi, A.K. Elhakeem / Computers and Electrical Engineering 36 (2010) 978–992 979

selects a potential Sponsoring Node to connect to. A Mesh Network Entry message (MSH-NENT) with NetEntryRequest infor-mation is then sent by the Candidate Node to join the mesh [6].

In this paper we introduce the transceiver and access techniques of a new CDMA/TDD based system and analyze the per-formance. The new network use the parallel transmission from the SS, use turbo coding and adaptation of SS transmission tothe network load (which affects queue control) are all combined to improve the network QoS. Moreover we make compar-isons with the TDMA based systems in terms of network efficiency, delay and delay jitter. Results show that CDMA systemout performs the TDMA counterparts.

2. System model

Fig. 1 shows the TDD operation of the proposed system. When the node (SS) powers ON, it is in passive mode and listensto HELLO message from nearby nodes and the lowest level in all HELLO message is j, it decides its own level as (j + 1). Forexample U5 receives HELLO messages from U2 and U6 which indicate they are in level 1 and level 3, U5 decides its levelto be 2.

In TDD mode nodes such as U1, U2, U3 in level 1, nodes U6, U7, in level 3, nodes U10 in level 5 all transmit in even num-bered slots 0, 2, 4, 6, . . ., while nodes BS and U4, U5 in level 2, node U8, U9 in level 4 all listen at same even slots. In odd slotsthe situation reverses, i.e. U1, U2, U3, U6, U7 and U10 all listen while BS, U4, U5, U8, U9 and U11 all transmit. Needless to sayupon power on and listening to HELLO messages and determining the smallest heard level as before the node determineswhich slot is odd or even.

For proper TDD operation however, reception slot is slightly enlarged by the maximum one way propagation delay of ahop.

3. Transmitter and receiver model

Fig. 2 shows a network scenario with multiple SSs and multiple end users. The mesh nodes (SS) consists of a transmitterand a receiver where the modulation and demodulation is performed. The synchronization between each mesh node and themobile end users are maintained by the base station pilot signal. The cellular base stations can perform this task as in 4G. It is

Fig. 1. A typical TDD mesh network.

Fig. 2. A network scenario with multiple SSs and multiple End users.

980 I. Al-wazedi, A.K. Elhakeem / Computers and Electrical Engineering 36 (2010) 978–992

foreseen that mesh nodes (SS) will be interconnected with the cellular base stations (Internetworking UMTS-mesh networks[29]). The immediate destinations of nodes packets are determined by the route decisions which are exchanged via the con-trol packets. Source destination fields as well as the identities of the Walsh functions of the subject sending node are insertedin the headers of transmitted packets. This will enable each receiving node to tune itself to the corresponding Walsh func-tions of this nearby transmitting node. Thus each receiving node will demodulate only the intended packets passing by thisnode as dictated by the routing policy (not all packets heard). This means that each node is actually a crossing road (switch)and it handles many data from many source destination pair at the same time. This implies that each node should have manyparallel demodulation banks working at same time and can process many packets of many neighboring nodes simulta-neously. The intended receiving node demodulates only the intended packets to be routed by this node. Nodes do not nec-essarily route all packets they hear from their neighbors. From the routing table, each node determines which neighboringpacket it should route and it configures its receiver bank and demodulates accordingly. It demodulates only the intendedpackets of specific neighbors and treats the rest of packets on the air as interference.

For the end users our protocol can be very efficient as each mesh node can modulate and demodulate many packets inparallel. For example, an end user is browsing a document from the internet and suddenly he receives voice or video. In thiscase our approach would be helpful for the users to accomplish both tasks simultaneously. Meaning multitasking is possibleusing our CDMA/TDD protocol.

The subject network is packet switched and not circuit switched. It is connection oriented in the sense that SSs have tosignal their destination before transmitting data packets. In this regard a priori establishment of the routes via exchangingcontrol packets is essential. Many routing techniques have been designed for routing packets among peer nodes in ad-hocWLANs [11–17]. Most of these are applicable to mesh networks and are actually traditional shortest path routing operatingin connection oriented mode. For rapid fault tolerance and autonomous load balancing other techniques can also be used[26–28]. In this paper we only analyze the system performance considering the interaction between the MAC and PHY layers,while for the network layer one of the efficient load balancing routing mechanism stated above can be used [7–10].

Fig. 3 shows the block diagram of the transmitter. Higher layer data and control packets are modulated by the spreadingcodes. In this work only one pilot code is used for all SSs for data and another pilot code for control as shown in Fig. 6. Alsoone longer scrambling code Cs is used by all transmitting nodes but with different shifts to scramble the whole of the datapacket except the pilot part as in Fig. 5. Another longer scrambling code C0s is used by all transmitting nodes but with dif-ferent shifts to scramble the control packet except the pilot part as in Fig. 6. The data packets from subject node or routed

Build Control Packets

Build Data Packets

Node Data From higher layers

Identities of Destination and Intermediate Nodes

Intermediate Packets to be routed

Information from receiver section

Configure switches based on TDD Operation and Node activity

Configure spread code shifts to selected destination or intermediate nodes

Control Information from receiver section

Carrier Stage, Up conversion

and Transmission

Data Code Cd

Fig. 3. Block diagram of the transmitter.


packets from intermediate nodes are spread by the pilot codes then TDD gated according to the TDD operation and the nodeactivities.

Fig. 4 shows the block diagram of the physical layer constituents of the receiver. The received signal is fed through afilter matched to data pilot code spreading. If the signal from the matched filter exceeds the threshold, this signal is as-signed a data bank and the data is demodulated by the data scrambling code Cs. If the signal does not exceed the thresholdwe continue to feed it through the matched filter until a threshold is detected. After demodulation of the data information,the header is checked using the CRC for the same subject node. If a preamble is detected, this bank is kept and the packetis fed to higher layer and also to the transmitter to initiate the data pilot spreading code. If no preamble is detected, wecheck for possible preambles from other nodes. In this case if a new preamble is detected, another demodulation bank isassigned. If no preamble is detected, the signal is discarded and corresponding bank is left for future use. The control sec-tion of Fig. 4 precedes the same as the data section but the signal is demodulated by the pilot spreading code for a smallpart of the control packet (Fig. 6) and by C0s. Figs. 5 and 6 show, respectively, the physical data and the physical controlpacket format. In addition to the short pilot spreading codes constituting the SYNC field of length 256 chips we also em-ploy a longer scrambling code of length ð215 � 1ÞTc . The starting phase is selected randomly by each user, same for all itspackets. Each packet starts its Cs from a certain initial condition. The initial state (starting phase of scrambling code) isconveyed by the field Cddi of Fig. 5 which is used by the receiver to be able to synchronize its local scrambling code tothat of the received packet. These di bits are spread by the short pilot data code for data packets and spread by the shortpilot control code for control packets.

The field dw of Fig. 6 conveys to the receiver, the current number of Walsh functions, i.e. the number of packets it is trans-mitting parallel in time. These dw bits are spread by the short pilot data spreading code.

Wi = ith Walsh function used i = 1, 2, 3, 4. A typical user employs a number of Np Walsh functions at receiver. dw bits arethe binary representation of Np. We do not transmit the identities of Walsh function used by each packet, such identitiesare assumed same to all nodes, e.g. if Np = 3, means Walsh function 1, 2, 3 are used by this packet.WL = Walsh function length = Tb see, after indicating the SYNC is detected, users employ Cd to detect, i.e. di bits the codestarting bits initial LFSR shift of long Cs code and number of Walsh function, i.e. Np.dw = bits indicate number of Walsh function i.e. number of packets transmitted in parallel.

LPF

LPF

r(t)

tc1cosω

tc1sinω

Data sync Matched filter

Threshold Exceeded

Continue Feed through MF

No

CS

Code Arrival time is a new code epoch for data packet

No

Discard

Adjust local such code as per detected code delay 1τ

Yes

Information Data Demodulated

Assign another datademodulation Bank 1

Preamble Detect, CRC Check for same subject node

Keep Bank1 feed Acquisition packet to higher layer

To higher layer

Yes

Discard and leave Bank 1 for future

No

1τ , select ID,

Hop number

TX

New Preamble Detect, CRC Check for same different node

Yes Assign another data demodulation Bank 1

Data Bank 1

Data Bank 2

Assign another data demodulation Bank n

Data Bank n

Control Sync Matched filter

Threshold Exceeded

Continue Feed through MF

No

YesCode Arrival time is a new code epoch for Control packet

No

Discard

Adjust local such code as per detected code delay 1τ

Yes

Information Data Demodulated

Assign another Control demodulation Bank 1

Control Bank 1

Preamble Detect, CRC Check for same subject node

Keep Bank1 feed Acquisition packet to higher layer

To higher layer

Yes

Discard and leave Bank 1 for future

No

1τ , select ID,

Hop number

TX

New Preamble Detect, CRC Check for same different node

Yes

Assign another Control demodulation Bank 1

Control Bank 2

Assign another Control demodulation Bank n

Control Bank n

No

CS

C�S C�S

Fig. 4. Block diagram of the receiver.


HeaderInformation BitsCRC Short pilot data spreading code

∑=

pN

jjjs dWC

1 sdhCW1 id dC wd dC dC

bT16 ( ) bb TN 100− bT56 bT15 bT15bT8

Routing and other bits 20bits

Piggyback ACK 12 bits

Sequence Number 12 bits

Virtual Connection (Label) 12 bits

chipsTb 32=

Fig. 5. Physical data packet format.

redaeHControl BitsCRC

∑=

pN

jjjs dWC

1 sdhCW1 ipdC wpdC pC

bT16 ( ) bb TN 100− bT56 bT15 bT15bT8

Control messages as per standards for example

chipsTb 32=

Short pilot Control spreading code

chipsCp 256=

Fig. 6. Physical control packet format.


di = bits indicate initial shift of Cs code.Cd = short data pilot code. We note that the beginning of each packet is spread by the short pilot code while the headerand data bits are spread by Walsh function and longer scrambling code.

4. Performance analysis

As discussed in Section 3, the CDMA system is capable of transmitting packets in parallel. So for analysis purposes weconsider a bulk arrival bulk service system shown in Fig. 7 for packets in queue of a typical node.

From Fig. 7 the state transition probabilities which are function of the packet arrival al and service probabilities bl aregiven below:

aij ¼ 0; i � j; i� j � B ð1Þaij ¼ 0; i � j; j� i � B ð2Þ

aij ¼XminðJþB�i;BÞ

l¼0

albiþl�j; i � j; i� j � B ð3Þ

aij ¼ a0 þXB

l¼1

albl; i ¼ j ¼ 0 ð4Þ

aij ¼XB

l¼0

albl; i ¼ j – 0 ð5Þ

aij ¼XB�ðj�iÞ

l¼0

ajþl�ibl; i � j; jj� ij � B ð6Þ

In this analysis K P BU; B is the maximum number of packets simultaneously sent in parallel from a typical node (B = 4). Uis the maximum number of users per hop (i.e. it is three in first hop of Fig. 1) which are taken as 5, 6 and 7 in this work.

40a

0 1 K2 3 4

00a

10a

01a

11a

20a

30a

31a

22a

12a

02a

03a

21a

23a 34a

33a44a

41a

42a

04a

32a 43a

14a

13a24a

15a 25a 35a

Fig. 7. State transition diagram of bulk arrival bulk service switching node underlay the TDD/CDMA label switched network.


In Fig. 7 each node, i.e. a SS or logically a switching node receives packet from other nodes to forward (switch action) thatadds to the node’s intrinsic packet generation (self-traffic). Assuming that during each packet time the user possible gener-ation probabilities are:

a1 ¼ a2 ¼ a3 ¼ a4 ¼ p=B; a0 ¼ 1� p

We assume that the simultaneous generation of one or more packets by each node is equally likely. Here p is a given usercombined packet generation parameter i.e. probability of generation at any packet time.

The intended node probabilities of success of parallel transmission of one or two or B packets on the channel areb1; b2; b3; . . . ; bB. These probabilities are obtained by averaging over the conditional transmission from other users; i.e.

bk ¼XBU

n¼k

unPk

XB

u2¼0u3¼0

XB

u2þu3¼n�k

Pu2 Pu3

!ð7Þ

Pk is the probability that the intended user transmits k packets in parallel (to be stated shortly after) Pu2 is the probabilitythat another user transmits u2 packets in parallel. Pu3 is similarly defined and in Eq. (7) we average over all u2, u3 values suchthat k + u2 + u3 + n. Here un ¼ ð1� Pbn Þ

Nb ; where un is the probability of packet success, n is the total number of parallelpackets transmitted on the channel, Nb is the number of bits/packet and k is the number of packets transmitting in parallel(1 or 2 or 3 or 4) and Pbn is the probability of bit error given n parallel packets on the channel. For convenience, Eq. (7) is forthree users. For example if the total number parallel successful packets is six (probability of success = u6), and if the intendeduser transmit two packets in parallel (Pk ¼ P2), other 4 packets must come from the other two users (2 + u2 + u3 = 6).

For different number of packets n, Eq. (8) will be used to find the values of SNR and subsequently the values of Pb from theturbo coding results of [21]. We assume a simple turbo code of rate ½ and 256� 256 interleaving.

ðSNRnÞ�1 ¼ 1EbN

þ n� 1PG

" #ð8Þ

Here n = 1, 2, 3, . . ., BU and Eb/N is considered for the thermal noise which is assumed as 0.001. Eq. (8) models the cumulativeeffects of n spread packets received in parallel at the intended user as an equivalent AWGN with interface density equal toS(n � 1)/W; where W is the spread spectrum bandwidth;

PG ¼ Tb ðBit durationÞTc ðChip durationÞ ¼WTb

The probability of a typical node transmitting one or more packets on the channel is dependent on his queue contentwhich is inherently a function of generation probability p and the probability of packet success on the channel i.e.b1; b2; b3; . . . ; bB.


P0 ¼ probability of node transmitting no packet on the channel ¼ nð0Þ ð9Þ

P1 ¼ probability of node transmitting one packet on the channel ¼ nð1Þ ð10Þ

Meaning switching node will transmit one packet if they have one packet in the queue.

PB ¼ probability of node transmitting B packets on the channel ¼ 1�XB�1

i¼0

nðiÞ ð11Þ

Meaning nodes will transmit B packet in parallel if there are B or more packets in the queue.P0; P1; P2; P3; P4 refers the transmission action of a switched SS. As the network configuration changes, so will the queue

distribution of n, hence P0; P1; P2; P3; P4 are dynamically adjusted by the node every packet so as to react to networkconfiguration.

nð0Þ; nð1Þ; . . . ; nðBUÞ are the steady state distribution of the values of packets in the queue.Starting with a certain estimated distribution for the number of packets in the queue nð0Þ; nð1Þ; . . . ; nðBUÞ values of

P0; P1; P2; P3; P4 can be computed from Eqs. (9)–(11). Based on corresponding values of P0; P1; P2; P3; P4 all al and0 bl are com-puted from Eqs. (1)–(6). Next we solve the state equation n ¼ ½A�½n�. Where aij are the elements of A and we obtain a solutionn. This is again used iteratively to get new values of P0; P1; P2; P3; P4 and new values of aij then a new solution n and the pro-cess repeats till the steady state i.e. values of n do not change iteration to iteration. To find the steady state solution of thenumber of packets at the intended user generation, i.e. nð0Þ; nð1Þ; . . . ; nðUBÞ, we initially take estimated values such as 0.5, 0.2,0.1, 0.1, 0.1.

For queue stability we require the following condition

Bþ 12

� �p � �b

where ð1� pÞ0þ 1ðpBÞ þ 2ðpBÞ þ � � � ¼ ðBþ12 Þp is the average sum of number of packets generated or routed. �b is the average

packet service probability given by:

�b ¼ 1b1 þ 2b2 þ � � � þ BbB

Since our network is connection oriented, call establishment time is a crucial measure. The call establishment time isdetermined by the sum of the buffer occupancy of all SS on the most crowded route taken by call establishment control pack-ets. To determine this, one has to work at the statistics of this most crowded request link.

Fig. 8 shows such a network having a source and a destination with few intermediate nodes. For the call establishment (asan worst case analysis) we work with the most crowded request link in Fig. 8 which is the path having three links(lpath number; link number). Since the selection of best route is done based on call establishment phase hence for data packetswe can assume the least crowded path.

From Fig. 8, to determine the maximum or minimum delay required for the most and least crowded request links, it isnecessary to find the probability of sum of packets in queue on the ith path having mi links has ri number of packets, whichcan be written as:

bi;ri ;mi¼Xk

li;1¼0

Xk

li;2¼0

� � �X li;mi¼0

li;1 þ li;2 � � � þ li;mi¼ ri

kYmi

j¼1

ni;j

where ni;j = probability of jth queue of ith path having ri packets.

Destination NodeSource Node

( )311,1 =ml

( )22,2 2=ml m

Intermediate Nodes

( )221,2 =ml

( )13,3 3=ml m

( )312,1 =ml

( )31,1 1=ml m

Fig. 8. Network scenario of the longest and shortest route.

Table 1The parameters used for the analysis.

Bandwidth(W)

Number of bits/packet(Nb)

Bit rate(Rb)

Processing gain(PG)

Turbo code rate(r)

Rate packets/s One packet in ms

32 M 1000 and 2000 1 M 32 (PG = W/Rb) 1/2 32�106 bits=s1000 bits=packet ¼ 32;000

132;000 ¼ 0:03125 ms


In Table 2 a typical distribution of packets in different paths is shown which is used to calculate the maximum delay forthe control packets. In this case, for convenience three paths ðnp ¼ 3Þ have been chosen. Where np is the value of the numberof paths available. From Table 2 a generalized form for different number of packets in the most crowded path (worst casecontrol packets) can be written as:

a0 ¼Qnp

a¼1ba;0;ma which multiply the probability of paths, each with zero packets to give the probability that the mostcrowded path has zero packets

Table 2Distribu

ri ¼

ri ¼

ri ¼

an ¼Xnp

a¼1

ba;n;ma

Ynp

i ¼ 1i–a

Xn�1

l¼0

bi;l;miþXnp

a¼1

Xnp

c ¼ 1c � a

ba;n;ma bc;n;mc

Ynp

i ¼ 1i–a–c

Xn�1

l¼0

bi;l;miþYnp

a¼1

ba;n;ma ; n � 0

an the probability that the most crowded path taken by the control packets has a sum of n packets in all its queues. First termof an corresponds to the intended crowded path has n packets on all queues, while sum of all packets in other paths is lessthan n. For the second term of an, two paths can have the same sum of packets on all queues, i.e. n while the third one has lessthan n. n = 1, 2, . . ., lmaxBU; where lmax is the maximum number of links in one path.

an actually depends on queue values of mi and np, but we dropped the subscripts for convenience. It follows that the prob-ability of finite number of packets (n) averaged over a certain number of paths (np) each having finite number of links (mi) isgiven by the following equation:

�hn ¼X

np

Xm1

Xm2

� � �X

mi

anjnp ;m1 ;m2 ;...;mi� pðm1;m2; . . . ;miÞ � pðnpÞ

where p(m1, m2, . . ., mi) = 1Pm1

Pm2 ��P

mi; is the probability of uniformly distributed m1, m2, . . ., mi links on the 1st, 2nd, . . .,

ith paths, and p(np) = 1/(Total possible number of paths npðmaxÞ); is the probability of the uniformly distributed 1st, 2nd, . . .,ith paths, respectively. The average delay for the call establishment, i.e. maximum sum of packets in all queues for the most

crowded path becomes: �k ¼PlmaxBU

n¼1 n�hn.The delay jitter for the control packets can be expressed as follows:

rh ¼ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiX

n2�hn � ð�kÞ2q

Now for the data packets, which commences after call establishment, the shortest path have been discovered. So the sta-tistics of the least crowded path should be followed by all data packets. A generalized form for different number of packets inthe least crowded path (data packets) can be written as:

cn ¼Xnp

a¼1

ba;n;ma

Ynp

i ¼ 1i–a

1�Xn�1

l¼0

bi;l;mi

!þXnp

a¼1

Xnp

c ¼ 1c � a

ba;n;ma bc;n;mc

Ynp

i ¼ 1i–a–c

1�Xn�1

l¼0

bi;l;mi

!þYnp

a¼1

ba;n;ma

where the first term of cn corresponds to the intended least crowded path having n packets on all queues while the sum of allpackets in other paths is greater than n. For the second term of cn, two paths can have the same sum of packets on all queues,i.e. n while the third one has greater than n. The third term is the multiplication of the probability of paths, each with n pack-ets to give the probability that the least crowded path has n packets.

tion of packets in different paths that corresponds to the choice of path 1 is the most crowded path.

0 ri at path 1 ri at path 2 ri at path 30 0 0

1 ri at path 1 ri at path 2 ri at path 3 ri at path 1 ri at path 2 ri at path 3 ri at path 1 ri at path 2 ri at path 31 0 0 1 1 0 1 1 1

2 2 0 0 2 2 0 2 2 22 1 0 2 2 1 � – –2 0 1 – – – – – –2 1 1 – – – – – –

0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45 0.50

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

Probability of user generating a new packet while the call is active (N b=1000)

Net

wor

k ef

ficie

ncy

for

five

Hop

s

Number of users=5 (CDMA)

Number of users=6 (CDMA)

Number of users=7 (CDMA)Number of users=5 (TDMA)

Number of users=6 (TDMA)

Number of users=7 (TDMA)

Fig. 9. Network efficiency for TDMA and CDMA systems.


The probability of finite number of packets (n) averaged over a certain number of paths (np) each having finite number oflinks (mi) is given by the following equation:

�en ¼X

np

Xm1

Xm2

� � �X

mi

cnjnp ;m1 ;m2 ;...;mi� pðm1;m2; . . . ;miÞ � pðnpÞ

The average delay for the data packets i.e. minimum number of packets in all queues of least crowded path and the delayjitter are expressed in Eqs. (12) and (13).

D ¼X

n�en ð12Þ

re ¼ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiX

n2�en � ðDÞ2q

ð13Þ

5. Results and discussion

In a CDMA system the network efficiency is given by Eq. (14) which is to be compared with Eq. (15) the network efficiencyfor the TDMA system [20].

gCDMA ¼U

2W� �b � Rb � r � number of hops � Nb � Overhead

Nb

� �ð14Þ

Here U�bRb is the successful number of bits/s for all users; r is the FEC rate which reduces the bandwidth efficiency by ½,number of hops is considered for the automatic band reuse, Nb�Overhead

Nb

� �is considered for the overhead effects and W is

the available spread spectrum bandwidth. Due to half duplex TDD operation we multiply ½. The efficiency of a TDMA systemcarrying the same traffic as the CDMA system while ignoring the channel errors, overhead bits and assuming minimum FEC.

gTDMA ¼U � number of hops � �b � Rb

2Wð15Þ

Eq. (14) implies the dependence of gCDMA on the channel conditions and correct packet detection probability through �b.This takes into account the certain worst case assumptions where the number of nodes and load increases. In such case prob-ability of correct packet detection will deteriorate with parallel transmission permitted. However even with this deteriora-tion CDMA still outperforms TDMA counterparts shown in Fig. 11.

Fig. 9 show the network efficiencies of CDMA and TDMA systems for different values of p (i.e. users generating a newpacket while the call is active) for five hops. It is evident that for this number of hops the network efficiency for TDMAout performs CDMA.

Fig. 10 shows the same comparison of CDMA and TDMA systems for different number of hops at a constant p. Here it isclear that, while giving much advantage (i.e. not considering overheads, channel errors, FEC (bandwidth efficiency reductioneffect)) to TDMA system, still CDMA system out performs TDMA counterparts.

Nb=1000, p=0.1, PG=W/Rb=32, Overhead=100 bits

-0.2

0

0.2

0.4

0.6

0.8

1

1.2

0 5 10 15 20 25

Number of Hops

Net

wo

rk e

ffic

ien

cy CDMA, Users=5

CDMA, Users=6

CDMA, Users=7

TDMA, Users=5

TDMA, Users=6

TDMA, Users=7



0

2

4

6

8

10

12

14

16

18

0 5 10 15 20 25

Number of Hops

En

d t

o E

nd

Del

ay in

Pac

kets

CDMA, Users=5

CDMA, Users=6

CDMA, Users=7

TDMA, Users=5

TDMA, Users=6

TDMA, Users=7

Fig. 11. End to end delay for TDMA and CDMA systems.


The average packet delay of CDMA system (Eq. (12)) is compared with the TDMA system given by the following well-known equation [20]:

DelayTDMA ¼U0 � gTDMA

2ð1� gTDMAÞþ U0

2þ 1 Packets

where the effective carried users U0 ¼ 2�U��b�number of hops�RbW .

The delay jitter of the CDMA system (Eq. (13)) is compared with the TDMA system as given by the following equation:

ri ¼

ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiXBU

i¼1

i2Pi � ð�dÞ2vuut

where �d ¼PBU

i¼1iPi and Pi ¼ ð1� qTDMAÞqiTDMA where qTDMA ¼ gTDMA is the load on the system (arrival rate/service rate). In this

case we take M/M/1 queuing [20] as a worst case scenario.From the delay values for both cases in Fig. 11, it is evident that when the number of hops increases, the delay increases

faster in the case of TDMA. Also as we increase the number of hops the efficiency of CDMA system improves and may exceed1 due to automatic band reuse while for TDMA, as the number of hops increases gTDMA will increase and it will reach to only1. This leads to an unstable situation yielding very high delays. TDMA problems culminate when the total number of users inall hops of the CDMA system will have to share a large single TDMA frame. Besides large delays peak transmission power willthen increase tremendously in TDMA systems compared to CDMA counterpart which is a well-known fact in wireless trans-mission. It is however possible in principle to divide the total number of users from all hops over many TDMA frames eachframe for a specific geographic region. In such case gateways will still be needed to relay the packet between different TDMAframes. This entails accumulating complete packets, subsequent queuing. Meaning the end to end TDMA delay will arise


from the individual TDMA frames (sub-networks) as well as delays at the gateway making TDMA delay even worse than thesingle frame for all case. Also needless to say TDMA access needs much higher transmission power than CDMA counterparts).FDMA based WIMAX system have synchronization, power, etc. constraints similar to TDMA system and so will perform clo-sely the same.

In Ref. [30], the performance analysis of a multi hop mesh network is proposed where the average delay and delay jitter iscalculated for a three-hop wireless path. The average delay for such three-hop mesh is 8.4 ms which is much higher than theaverage delay (CDMA) shown in Fig. 11. Also our delay jitter (Fig. 12) performance for CDMA system outperforms the delayjitter calculated in [30] (where the delay jitter is 135.5 ms). In this paper the delay and delay jitter performance is calculatedin packets. However, translation from packets to milliseconds is given in Table 1.

Information bit duration in CDMA is considered as (1/Rb) i.e. 1 ls. While for the equivalent TDMA, it compresses to 1/W;where W is the total available bandwidth 32 M in this analysis. In the course of comparison between CDMA and TDMA WI-MAX we have put TDMA at relative advantages for example we ignored the channel errors, overhead bits and had no FEC rateloss. Yet CDMA delay performance is better.

In Ref. [31], the delay performance for wireless mesh network (collision-free MAC) was calculated as 2.5 ms at 40 packets/s which translate into a packet generation probability of p = 0.1. At p = 0.1 our CDMA delay performance (Fig. 11) outper-forms such delay in [31] at comparable number of hops, p value, etc.

Fig. 13 compares the end to end delay jitter for both CDMA and TDMA systems. It is evident that the delay jitter increasesand the system become unstable for higher values of p, for the TDMA system. While at higher value of p, the CDMA systemcan still be operated.

Fig. 14 shows a comparison of the network efficiency between the two systems for different number of bits/packet. It canbe seen that there is not much variation in the network efficiency in the case of TDMA system since overhead is neglected.While for the CDMA system the network efficiency varies widely due to overhead changes.

Fig. 15 shows the end to end delay performances of the two systems for different number of bits/packet. For the CDMAcase the delay does not change much. The effect of number of bits/packet is less on the probability of successful packet


0

2

4

6

8

10

12

14

16

18

0 5 10 15 20 25Number of hops

En

d t

o E

nd

Del

ay J

itte

r in

P

acke

ts CDMA, User=5

CDMA, User=6

CDMA, User=7

TDMA, User=5

TDMA, User=6

CDMA, User=7

Fig. 12. End to end delay jitter for TDMA and CDMA systems.

Nb=1000, PG=W/Rb=32, Overhead=100 bits

0

2

4

6

8

10

12

14

16

18

20

0 5 10 15 20 25

Number of Hops

En

d t

o E

nd

Del

ay J

itte

r in

P

acke

ts

CDMA,Users=5,p=0.1

TDMA,Users=5,p=0.1

CDMA,Users=5,p=0.27

TDMA,Users=5,p=0.27

CDMA,Users=5,p=0.37

TDMA,Users=5,p=0.37


p=0.1, PG=W/Rb=32, Overhead=100 bits

0

2

4

6

8

10

12

14

16

18

0 5 10 15 20 25

Number of Hops

En

d t

o E

nd

Del

ay in

Pac

kets

CDMA,Nb=1000,Users=7

TDMA,Nb=1000,Users=7

CDMA, Nb=2000,Users=7




-0.2

0

0.2

0.4

0.6

0.8

1

1.2

0 5 10 15 20 25

Number of Hops

Net

wo

rk e

ffic

ien

cy







transmission on the channel. From this figure it is evident that Delay performance is much better in the case of CDMAsystem.

Fig. 16 shows the delay jitter for the systems at low value of p. It is evident that at low value of p, the number of bits/packet has little effect.


0

2

4

6

8

10

12

14

16

18

0 5 10 15 20 25

Number of hops

En

d t

o E

nd

Del

ay J

itte

r in

P

acke

ts

CDMA,Nb=1000,User=7

TDMA,Nb=1000,User=7

CDMA,Nb=2000,User=7

TDMA,Nb=2000,User=7

Fig. 16. End to end delay Jitter for TDMA and CDMA systems.

PG=W/Rb=32, Overhead=100 bits

-1

4

9

14

19

24

0 5 10 15 20

Number of Hops

En

d t

o E

nd

Del

ay in

Pac

kets CDMA,

User=5,Nb=1000,p=0.37

TDMA,User=5,Nb=1000,p=0.37

CDMA,User=5,Nb=2000,p=0.37

TDMA,User=5,Nb=2000,p=0.37


PG=W/Rb=32, Overhead=100 bits

0

2

4

6

8

10

12

14

16

18

20

0 10 20 30

Number of Hops

En

d t

o E

nd

De

lay

Jit

ter

in

Pa

cke

ts

CDMA,Users=5,Nb=1000, p=0.37

TDMA,Users=5,Nb=1000, p=0.37

CDMA,Users=5,Nb=2000, p=0.37

TDMA,Users=5,Nb=1000, p=0.37



Fig. 17 shows a comparison of the two systems at higher value of p. The variation in this case is visible. This means thatthe value of p (i.e. users generating a new packet while the call is active) affects the two systems. The system becomes unsta-ble more noticeably in the case of the TDMA system.

Fig. 18 shows the comparison of the delay jitter for the two cases. Again it is evident that the delay jitter performance getsworse as p increases for both cases. The TDMA system gets unstable more rapidly at higher value of p.

6. Conclusion

In this paper a new CDMA/TDD system has been proposed for wireless mesh networks. This was followed by a compar-ison between CDMA/TDD and TDMA/TDD based networks. The system building blocks for both control and data planes wereintroduced. Next Qos criteria were evaluated for both control and data packets. Analysis shows that the versatility of CDMAapproach is better suited to larger number of hops (i.e. large number of SS) compared to the TDMA counterpart in terms ofnetwork efficiency, delay, and delay jitter. Moreover CDMA is a plug and play, i.e. instantly reconfigured operation with lowpower consumption which is an important asset towards mesh networks.

References

[1] Yun Jungnam, Kavehrad Mohsen. PHY/MAC Cross-layer Issues in Mobile WiMax. Bechtel Corporation; 2006.[2] IEEE Std. 802.16-2004: IEEE standard for local and metropolitan area networks part 16: air interface for fixed broadband wireless access systems; 2004.[3] IEEE P802.16-Cor1/D5, Draft corrigendum to ieee standard for local and metropolitan area networks – part 16: air interface for fixed broadband

wireless access systems; 2005.[4] IEEE P802.16e/D11, Draft amendment to IEEE standard for local and metropolitan area networks – part 16: air interface for fixed and mobile broadband

wireless access systems – amendment for physical and medium access control layers for combined fixed and mobile operation in licensed bands; 2005.[5] TDD coalition white paper, The Advantages and benefits of TDD broadband wireless access systems; 2001.[6] Wei Hung-Yu, Ganguly Samrat, Izmailov Rauf, Haas Zygmunt J. Interference-aware IEEE 802.16 WiMax mesh networks. In: 61st IEEE vehicular

technology conference (VTC 2005 spring), Stockholm, Sweden; 2005.


[7] Perkins C. Ad-hoc on-demand distance vector routing. In: MILCOM, ’97 panel on ad hoc networks; 1997.[8] Clausen T, Jacquet, P. Optimized link state routing protocol, IETF internet draft, draft-ietf-manet-olsr-11.txt; 2003.[9] Michalak M, Braun T. Common gateway architecture for mobile ad-hoc networks. In: WONS’05: Proceedings of the second annual conference on

wireless on-demand network systems and services (WONS’05). Washington, DC, USA: IEEE Computer Society; 2005. p. 70–5.[10] Rainer Baumann and Vincent Lenders and Simon Heimlicher and Martin May, ‘‘HEAT: Scalable Routing in Wireless Mesh Networks using Temperature

Fields,” in Proceedings of the IEEE International Symposium on a World of Wireless, Mobile and Multimedia Networks (WoWMoM), 2007.[11] Perkins CE, Royer EM. Ad-hoc on-demand distance vector routing. In: Proceedings of the IEEE workshop on mobile computer systems and applications;

1999. p. 90–100.[12] Karp B, Kung HT. GPSR: greedy perimeter stateless routing for wireless networks. In: Proceedings of MobiCom, Boston, MA, USA; 2000.[13] Joshi A et al. HWMP specification. IEEE P802.11 Wireless LANs, document IEEE 802.11-061778r1; 2006; Lenders V, May M, Plattner B. Density-based

vs. proximity-based any cast routing for mobile networks. In: Proceedings of the IEEE Infocom, Barcelona, Spain; 2006.[14] Tonguz OK, Ferrari G. Ad hoc wireless networks: a communication-theoretic perspective. John Wiley and Sons; 2005.[15] Bantz David F, Bauchot Frédéric J. Wireless LAN design alternatives. IEEE Network 1994;8(2):43–53.[16] Basagni S, Chlamtac I, Syrotiuk VR. Dynamic source routing for ad hoc networks using the global.[17] Positioning System. In Proceedings of the IEEE wireless communications and networking conference 1999 (WCNC’99); 1999.[20] Leon-Garcia Alberto, Widjaja Indra. Communication networks. Fundamental concepts and key architectures.[21] Berrou C, Glavieux A, Thitimajshima P. Near Shannon limit error-correcting coding and decoding: turbo-codes 1.[26] Jung Sangsu, Kserawi M, Lee Dujeong, Rhee J-KK. Distributed potential field based routing and autonomous load balancing for wireless mesh networks.

IEEE Commun Lett 2009;13(6):429–31.[27] Rong Bo, Qian Yi, Lu Kejie, Hu RQ. Enhanced QoS multicast routing in wireless mesh networks. IEEE Trans Wireless Commun 2008;7(6):2119–30.[28] Li Y, Yang Y, Cao C. A novel routing algorithm in distributed IEEE 802.16 mesh networks. IEEE Commun Lett 2009;13(10):761–3.[29] Wiethoelter Sven, Wolisz Adam. Selecting vertical handover candidates. In: IEEE 802.11 mesh networks, world of wireless, mobile and multimedia

networks and workshops, 2009, WoWMoM 2009. IEEE international symposium; 2009. p. 1–7.[30] Chen Yu, Chen Jia, Yang Yang. Multi-hop delay performance in wireless mesh networks. Int J Mobile Networks Appl 2008;13(1-2):60–168.[31] Wang Ping, Zhuang Weihua. A collision-free MAC scheme for multimedia wireless mesh backbone. IEEE Trans Wireless Commun 2009;8(7):1–2.

Imam Al-Wazedi graduated from Khulna University of Engineering and Technology, Khulna, Bangladesh in Electrical andElectronic Engineering. He finished his Master’s from Ritsumeikan University, Japan in Information Science and SystemsEngineering. He completed his Ph.D. in the Department of Electrical and Computer Engineering at Concordia University,Montreal, Canada. His research interest include cross layer design, mesh networks, etc.

Ahmed K. Elhakeem received the Ph.D. degree from the Southern Methodist University, Dallas, TX, in 1979. He spent the nexttwo years working as a Visiting Professor in Egypt, after which he moved to Ottawa, Canada, in 1982. He assumed research andteaching positions in Carleton and Manitoba Universities and later moved to Concordia University, Montreal, Canada, in 1983,where he is now a Professor in the Electrical and Computer Engineering Department. He has published numerous papers in IEEEand international journals in the areas of CDMA and networks, and contributed to many books. He is a well-known expert inthese areas and serves as a consultant to various companies. His current research interests include cross layer design, inter-connected wireless LANs, 4G, wideband networks, WCDMA code Acquisition, CDMA networks, and reconfigurable networks, Heis a co-author of the book, Fundamentals of Telecommunications Networks (New York: Wiley, 1994). He has chaired andorganized numerous technical sessions in IEEE conferences, and was the Technical Program Chairman for IEEE Montech’86 inMontreal, Canada. More recently, he was the Key Guest Editor for four issues of the IEEE JOURNAL ON SELECTEC AREAS INCOMMUNICATIONS on Code Division Multiple Access, namely CDMA I, II, III and IV, appearing in May ’94, June ’94, October ’96and December ’96. Dr. Elhakeem was the Communications chair of IEEE Montreal and TCCC Rep. to ICC ’99, He served as an

associate editor for the IEEE Communications letters Journal (1996–1999), he is a professional engineer of Ontario and a senior member of IEEE, and IEICE.
Dr. Elhakeem has served as a consultant to many companies, to Communication Research Center and Industry Canada.

Documents

1 s2.0-s0045790610000182-main