Advanced wireless networks 4g technologies phần 6 docx

Therefore, a joint control of data rate as well as the transmitted power is an importanttopic in modem communication systems.. Georganas, A hybrid channel assignment scheme in large scal

Offered traffic of the three service classes in 4:3:3 MTIP

Figure 12.11 MdCAC vs MsCAC

Offered traffic of the three service classes in 4:3:3 MTIP MEM: memory(auto-regressive) MsCAC

MLESS: memory less MsCAC

Figure 12.12 Memory vs memoryless MsCAC systems

that the offered traffic of class 1 is 7λ Erlangs, class 2 is 2λ Erlangs and class 3 is λ Erlangs.

Thus, the total offered traffic is 10 Erlangs ifλ = 1.

Figure 12.11 presents the new-call blocking as well as the handoff failure probability ofeach service class in MdCAC and MsCAC systems vs the common divisor of the offeredtraffic intensities of the service classes in 4:3:3 MTIP This figure confirms that there is

no capacity gain using MsCAC for serving CBR services Figure 12.12 shows the needfor stable and reliable measurements in MsCAC systems The performance characteristics

of a memoryless measurement-based and a memory (auto-regressive) measurement-basedsystem are presented also for 4:3:3 MTIP Estimations with the help of auto-regressivefilters may results in better performance, but more complex hardware/software is needed.Figures 12.13–12.15 present the handoff failure and new-call blocking probabilities of the

Offered traffic of the three service classes in 4:3:3 MTIP

Figure 12.13 SCAC vs MdCAC in 4:3:3 MTIP

Offered traffic of the three service classes in 5:4:1 MTIP

Figure 12.14 SCAC vs MdCAC in 5:4:1 MTIP

SCAC system in comparison with the MdCAC system for 4:3:3, 5:4:1 and 7:2:1 MTIPsrespectively Figure 12.16 demonstrates that the SCAC system with the incomplete Gammadecision function offers better average communication quality, but slightly worse equi-librium blocking and dropping characteristics compared with the Gaussian function Thefreedom of choosing such decision functions to fulfill the performance requirements in-creases the flexibility of the SCAC policy The 4:3:3 MTIP is used in Figure 12.16 TheQoS loss probability vs the offered traffic in 7:2:1 MTIP are listed in Table 12.6 for different

Figure 12.15 SCAC vs MdCAC in 7:2:1 MTIP.

Offered traffic of the three service classes in 4:3:3 MTIP

Figure 12.16 Flexibility of choosing decision function in the SCAC system

Table 12.6 QoS loss probability in comparisonThe three-class RT offered traffic, e.g in 7:2:1 MTIP

421

Figure 12.17 SCAC system with threshold hard blocking QoS differentiation.

CAC systems Although the SCAC system suffers slight degradation of the communicationquality, it yields significant improvements in the handoff failure probability and in thecall blocking probability The traffic shaping gain of the SCAC is clearly illustrated inFigure 12.15, where the traffic intensity of the voice calls is really high For a predefinedperformance requirements, e.g less than 0.1 and 0.5 % handoff failure probability for voiceand other calls respectively, less than 1, 5 and 10 % new-call blocking probability forclass 1, class 2 and class 3, respectively, and 10 % allowable equilibrium outage probability,the SCAC system overall offers much better Erlang capacity Moreover, there is no need forredesign of the capacity thresholds in the SCAC system as long as the range of allowableuncertainty is maintained with the help of other control mechanisms such as TPC, link-adaptation, etc Thus, the robustness is also well improved over the MdCAC system SCAChas demonstrated an efficient RRU and capacity enhancement

With QoS differentiation, operators can customize the operation of serving networks.Figure 12.17 presents the performance characteristics of the following simple scenario

Users are divided into two user classes: business ( j = 1) and economy class ( j = 2)

Re-quests of the business users for any RT services are served immediately as long as thereare enough resources for accommodating them On the other hand, requests of the econ-omy users are served only if less than 70 % of effective resources are occupied by RT

traffic, i.e system load state c less than 0.5 Assume that demands for services of user

classes are equal Thus, arrival rates of new call requests from user classes for each serviceclass are equal Invoke assumption (3b) withλ l ,1k = λ l ,2k for k = 1, 2, 3 The offered traffic

of e.g 4:3:3 MTIP above can be split for each user class resulting in 2:2:1.5:1.5:1.5:1.5

MTIP of six traffic classes The factor a 0 jk (c) of admission probability in (12.56) can

be determined by (12.49), where the blocking threshold of business class l 1k is Cu and

of economic class l 2k is 0.5 for all k This numerical example clearly demonstrates the

effects of QoS differentiation on performance characteristics The business class not onlyexperiences much better GoS, but also better communication quality during the calls

Figure 12.18 illustrates the QoS differentiation with load-based fracturing factors for blocking of new calls of the economic class This is based on a simple scenario as follows.Again we assume handoff calls have the highest priority regardless of associated user class.The business class is served as long as resources are available The economy class canshare the resources equally with the business class if less than 65 % of effective resources

soft-are occupied, i.e c is less than 0.47 Otherwise, if c is less than C l, invoke (12.56) with:

a0 21(c) = 0.8, a0 22(c) = a0 23(c) = 0.6 If c is less than Cu, a0 21(c) = 0.4, a0 22(c)=

a0 23(c) = 0.3 Otherwise, a0 21(c) = a0 22(c) = a0 23(c) = 0.

The offered traffic is the same as in the previous scenario Figures 12.17 and 12.18show that the performance characteristics of the system can easily be tuned by using eitherthreshold-based hard blocking or fracturing factor-based soft blocking paradigms For NRTpacket radio access, additional parameters are given in Table 12.7

The average upper-limit UL data throughput for packet transmissions with 64 kbs and

10 ms TTI is presented in Figure 12.19 vs different offered traffic intensities of the three

RT service classes, which are in 7:2:1, 5:4:1 and 4:3:3 MTIPs The impacts of bit-rates

Figure 12.20 Figure 12.21 illustrates the effects of Tpwith a constant bit-rate of 64 kbs.Table 12.8 summarizes the mean values of aggregate RT traffic and quasi-stationary free

Offered traffic of the three service classes in 4:3:3 MTIP BxCS: SCAC system without QoS differentiation BxxQF: with fracturing QoS differentiation

Figure 12.18 SCAC system with fracturing soft blocking QoS differentiation

Table 12.7 Parameter summary for packet radio access

bit-rates, respectively

0.5 1 1.5 2 2.5 3 3.5 4 0

5 10 15 20 25

Common divisor of the RT offered traffic

Figure 12.19 Average upper-limit UL data throughput in different RT MTIPs

0 5 10 15 20 25 30 35

Common divisor of the RT offered traffic

RT offered traffic in 7:2:1 MTIP

Figure 12.20 Effects of the bit-rates to the throughput

capacity over Tptime-interval of 10 ms Figure 12.20 and 12.21 and Table 12.8 are for 7:2:1MTIP One can see that throughput characteristics are affected significantly by the dynamic

of RT traffic as well as the packet-transmission parameters The results give valuable titative merits for studying the design parameters and the performance tradeoffs of packetaccess control schemes This explains the motivations of using DFIMA scheme presentedabove, where content of feedback information provides 1:1 mapping of optimal transportformat combination (including TPP, bit-rate, packet-length or TTI) for packet transmission

quan-in the next time-slot based on feasible free resource predictions For example, assume 50 %cell capacity is occupied by the RT traffic of the 7:2:1 MTIP at the end of a given time slot of

10 ms The possible bit-rates for packet transmissions are 32, 64 and 144 kbs Consider

0.5 1 1.5 2 2.5 3 3.5 4 2

4 6 8 10 12 14 16 18 20 22

Common divisor of the RT offered traffic

*Line: T p =30 ms oLine: T p =40 ms

Figure 12.21 Effects of the packet transmission durations

Table 12.8 Means of stationary RT aggregate traffic and quasi-stationary NRT resource

availabilityThe three-class RT offered traffic, e.g in 7:2:1 MTIP

expected throughput in this case is about eight packets since the free capacity is successfullyutilized for packet transmissions with the probability of 1/2 The performance of DFIMAcan be optimized with respects to TPP, bit-rate, TTI and QoS differentiation paradigms,which is flexible and effective

often-complex statistics, some of the parameters may be hard to determine without whichthe modeling-based CAC policies cannot operate The soft-decision solutions are believed

to give more flexibility in determining the modeling parameters, and thus are quite suited

to achieving good multiplexing gain and robustness On the other hand, implementations

of the MsCAC policy require advanced hardware and software to ensure the reliability ofmeasurements For this reason, it is not cost-effective Moreover, estimation errors in somecircumstances may cause significant degradations of the system performance However, theadvantage of MsCAC is that it seems ‘insensitive’ to the traffic nature and the operation isrobust The network can learn and adapt to the statistics of traffic even when the burstiness oftraffic is considered as out of control for the modeling-based systems To gain tradeoff of alldesign criteria, a hybrid soft-decision/measurement-based implementation is a reasonablechoice Parameters needed for soft-decision functions, i.e means and variances, can rely

on auto-regressive measurements For such solution, parameters and constraints can simply

be thresholds of the UL interference level of cell and connection basis, an allowable outageprobability, estimates of the current total received interference level with its mean andvariance, etc These are anyhow needed for the TPC mechanism of CDMA systems Forimplementations of the DFIMA scheme, measurements or estimations of RT system loadstate for CAC can be reused The NRT offered traffic needs to be measured or estimatedfor prediction of optimal parameters (e.g TPP, bit-rate, TTI) that are used as the content

of feedback information Look-up tables for transport format combinations of UL packettransmission can be implemented or configured in both mobile and access network sides inorder to minimize the size of feedback information Eight bit feedback is enough to ensuresufficient exchange of control information in DFIMA, even with QoS differentiation

12.5 JOINT DATA RATE AND POWER MANAGEMENT

As already seen from the previous discussion the radio resource manager (RRM) contains

a number of sub-blocks like the connection admission controller, the traffic classifier, theradio resource scheduler and the interference and noise measurements The main role ofthe RRM is to manage the different available resources to achieve a list of target QoS.The radio resource scheduler (RRS) is an essential part of the RRM The RRS has twoimportant radio resources to control: MS transmitting power and transmitted data rate.The RRS uses those two resources to achieve different objectives like maximizing thenumber of simultaneous users, reducing the total transmitting power, or increasing the totalthroughput The conventional way to achieve these objectives is to select one of them as

a target to optimize and use other objectives as constraints More sophisticated algorithmsbased on multiobjective (MO) optimization and Kalman filter techniques have been alsoproposed Here we address the problem of how to combine the power and rate in an optimumway

Even Shannon’s equation shows that the achievable information rate in a radio nel is an increasing function of the signal-to-interference and noise ratio Increasing theinformation rate in data communication systems is restricted by the SINR Increasingthe SINR can be done in two ways The first way is by reducing the total interferenceand noise affecting that user This depends on some characteristics of the noise and theinterference For example, if the structure of the interference from other users is known at

chan-the receiver chan-then, by applying one of chan-the multi-user detection methods, that interference can

be reduced Also if the users are spatially distributed then the interference can be reduced byusing a multi-antenna system (see Section 12.1) If the users concurrently use the channel(as in DS-CDMA) then the interference can be reduced using power control techniques.From previous studies we can see that some characteristics of the interference are assumed

to be known or can be controlled There are many sources of interference and noises thatcannot be reduced by the first way such as thermal noise, interference from other cells etc.The second way of increasing the SINR is simply by increasing the transmitted power In

a single user communication (point-to-point) or in broadcasting, this can be an acceptablesolution and the main disadvantages are the cost and the nonlinearities in the power am-plifiers However, in a multiuser communication environment, increasing the transmittedpower means more co-channel and cross-channel interference problems

Therefore, a joint control of data rate as well as the transmitted power is an importanttopic in modem communication systems The modern communication systems (3G or 4G)are supporting the multirate data communication because they are designed not only forvoice communication but also for data and multimedia communication

An efficient combining algorithm for the power control and the rate control is requiredfor these systems The term ‘efficient’ here refers to optimization of the transmitted powerand data rate to meet the required specifications There are many proposed combiningalgorithms for power and rate control in the literature The objectives of those algorithmsare quite varied Some algorithms suggest maximizing the throughput; others minimizingthe packet delay or minimizing the total power consumption

The 3G/4G mobile communication systems based on WCDMA support the rate transmission There are mainly two methods to achieve the multi-rate transmission,the multicode (MC) scheme and the variable-spreading length (VSL) scheme In theMC-CDMA system, all the data signals over the radio channel are transmitted at a ba-

multi-sic rate, Rb Any connection can only transmit at rates m Rb, referred to as m-rate, where m

is a positive integer When a terminal needs to transmit at m-rate, it converts its data stream, serial-to-parallel, into m basic-rate streams Then each stream is spread using different and

orthogonal codes In a VSL-CDMA system, the chip rate is fixed at a specified value (3.84Mb/s for UMTS) and the data rate can take different values This means that the processinggain (PG) is variable The processing gain can be defined as the number of chips per symbol

12.5.1 Centralized minimum total transmitted power (CMTTP) algorithm

[P1, , PQ]Tand the rate vector R = [R1, , RQ]Tminimizing the cost function:



j=1 P j G k j + N i

is called infeasible In this case either some user should be dropped from this link or some

of the constraints should be relaxed At the optimal solution, all QoS constraints are metwith equality Also, the optimal power vector is the one that achieves all rate constraints

with equality So, the optimum rate vector is R*= [r1, , rQ]T The corresponding powervector can be obtained by solving the QoS equation This is a system of linear equations inpower From Equation (12.69) we have

Rs

r i

P j G k j Q

i is the target SINR for user i Let ˜r i = δT

i r i /R s and substitute it into Equation(12.71), to obtain

In order to obtain a nonnegative solution of Equation (12.77), the following condition shouldhold:

ρ(rH) < 1

whereρ(A) is the spectral radius of matrix A.

12.5.2 Maximum throughput power control (MTPC)

This algorithm has been suggested in Chawla and Qiu [85] The algorithm is based on themaximization of the total throughput in a cellular system There is no need to generateall solutions in this method Since the gain links and the interference of other users are

needed to calculate the transmitted power of each user, the MTPC algorithm is a centralized

algorithm The throughput of user i can be approximated when M-QAM modulation is used

gain between mobile station j and base-station k The total throughput T is given by

where Q is the number of users.

Now the problem can be defined as follows: given the link gains G i j of the users, what

is the power vector P = [P1, P2, , PQ] which maximizes the total throughput? Since

the first term in Equation (12.77) is constant and the logarithmic function is an increasingfunction, then maximizing the multiplicative term

Equation (12.77a) is given by

where G i j is the channel gain between user j and base station i Without loss of generality

user i is assumed to be assigned to base station i Starting from any initial vector P(0) ∈ , the

iteration specified by Equation(12.78) converges to a unique point P*∈ , which achieves

the global maximum [85]

12.5.3 Statistically distributed multirate power control (SDMPC)

A distributed solution of the optimization problem given by Equations (12.68)–(12.70) is

proposed for one cell case in Morikawa et al [80] It is assumed that every user has two states, active ON or passive OFF The transition probabilities of the i th user from idle to

active state at any packet slot isυ i, and from active to idle state isζ i The durations ofthe active and idle periods are geometrically distributed with a mean of 1/ζ i and 1/υ i (inpacket slots), respectively The optimization problem Equations (12.68)–(12.70) is slightlymodified as

Markovian property since geometric distribution is memoryless over the duration of traffic.The centralized solution (if the system is feasible) is given by

j=1β j (t) γ j) is estimated The Markovian property of the randomprocessβ j (t) has been exploited to obtain a good estimate of the other users’ information

part The SDMPC algorithm is given by

active (when the mobile phone is ON) Then, in the SDMPC algorithm, the channel gainand the average power of the additive noise are assumed to be known In reality they should

be estimated as well Good estimation of the channel gain and the noise variance is usuallydifficult In practice it is easer to estimate CIR or SINR because they have direct impact onBER Finally, it was assumed that the durations of active and idle periods are geometricallydistributed This assumption is oversimplified and far from reality

12.5.4 Lagrangian multiplier power control (LRPC)

As mentioned previously, the data rates which can be achieved belong to a set of integers

In the formulation of the optimization problem, to maximize the data rate we assume thatthe date rate is continuous This assumption can be relaxed in the simulation by roundingthe optimum data rate to the nearest floor of the data rate set It can be proven that thesolution of the optimization problem with continuity assumption is not necessarily thesame as the solution of the actual discrete problem The advantage of the LRPC algorithm

is that the optimization problem has been formulated without the continuity assumption

of the data rates [81] It has been assumed that each user has a set of m transmission

rates M= {r1, r2, , rm} to choose from Let the rates be ordered in as ascending way,

i.e r1< r2 < < r m To properly receive messages at transmission rate r k , mobile i is

as the outage probability is rather high Therefore it is not recommended to be used in thesystems where the fairness is an important issue

12.5.5 Selective power control (SPC)

The SPC algorithm has been suggested in Kim et al [81] The SPC algorithm is a logical

extension of the DCPC algorithm [82] The main idea of the SPC algorithm is to adapt thetarget CIR of each user to utilize any available resources The suggested SPC algorithm isgiven by

whereχ(E) is the indicator function of the event E Although the SPC algorithm improves

the outage probability compared with LRPC algorithm, its outage is still high

J¨antti suggested an improved version of the SPC algorithm It is called selective powercontrol with active link protection (SPC-ALP) Algorithm [83] The SPC-ALP algorithmhas less outage probability and better performance than the SPC algorithm The main idea

of the SPC-ALP algorithm is to admit new users into the network with at least the minimumdata rate and also if possible allow old users to choose higher data rates This is done bydefining three different modes of operation for each user:

rate cannot be increased but it can be decreased if needed If there are more resources to

be utilized by increasing the rate, the used mode is changed to the transition mode

r Transition mode, where the user updates its power using ALP algorithm Also the rate isadapted to the maximum rate that can be supported

algorithm can be found in J¨antti and Kim [83]

12.5.6 RRM in multiobjective (MO) framework

The QoS can be defined for a set of factors In this Section we will consider only the BERand the user data rate in the uplink The objectives of the RRS could be defined as(1) Minimize the total transmitting power

(2) Achieve the target SINR in order to achieve a certain BER level (depends on theapplication)

(3) Maximize the fairness between the users In our definition, the system is fair aslong as each user is supported by at least its minimum required QoS In this sense,minimizing the outage probability leads to maximizing the fairness

(4) Maximize the total transmitted data rate or at least achieve the minimum requireddata rate

It is clear that objective (1) is totally conflicting with objective (4) and partially conflictingwith objective (2) Objective (3) is totally incompatible with objective (4) Objective (2) ispartially contradictory with objective (4)

So far, the RRM problem was formulated as a single objective (SO) optimization lem considering the other parameters as constraints Solving the objectives (1)–(4) at thesame time as using MO optimization technique, leads to a more general solution thanthe conventional methods In this section we discuss an MO optimization method to solvethe RRM problem In subsequent subsections we will discuss some radio resource sched-uler algorithms based on MO optimization The field is very wide and many differentalgorithms and methods can be derived based on MO optimization One formulation of theRRS optimization problem can be defined as:

Pmin ≤ P i ≤ Pmax, Ri ,min ≤ R i ≤ R i ,max (12.91)

where O P is the outage probability The outage probability is defined as the probabilitythat a user cannot achieve at least the minimum required QoS We can see that the O Preflects the fairness situation in the communication system The minus sign associated withthe sum of the rate function in Equation (12.90) refers to the maximization process of thetotal utility functions

Defining the objectives and the constraints is the first step Selecting the proper MOoptimization method to solve the problem is the second step Then the (weakly) Paretooptimal set of solutions is generated, where every solution is optimal in a different sense.Finally, the decision maker selects the optimum solution from the optimal set which bestachieves the required specifications In this section we discuss a framework to use the MOoptimization techniques in RRM

12.5.7 Multiobjective distributed power and rate control (MODPRC)

The algorithm is based on minimizing a multi-objective definition of an error function Inthis algorithm we defined three objectives: (1) minimize the transmitted power; (2) achieve

at least the minimum CIR, which is defined at the minimum data rate; and (3) achievethe maximum CIR, which is defined at maximum data rate An optimized solution can

be obtained using an MO optimization

The derivations of the algorithms are based on a VSL-CDMA communication system.After the dispreading process at the receiver, the SINR is

δ i (t)= Rs

R i (t)  i (t) , t = 0, 1, (12.92)whereδ i (t) is the SINR of user i at t, Rsis the fixed chip rate (= 3.84 Mb/s for UMTS),

R i (t) is the data rate for user i at t, and  i (t) is the CIR of user i at t In wireless and digital

communication, it is well known that the BER is a decreasing function in the SINR Incase of coherent binary PSK, the BER can be approximated by (when the interference isassumed Gaussian)

2erfc

'√

For example, if the BER should not be more than 10−4then the target SINR is obtained fromEquation (12.93) asδT ≥ 8.3 dB In the case of fixed data rate power control there is one

target CIR that corresponds to the target SINR, because we have only one spreading factorvalue In a case of multirate services there are different target CIR values corresponding

to the target SINR From Equation (12.92) it is clear that, in case of constant target SINR,maximizing CIR leads to maximizing data rate as follows:

The target SINR, the minimum/maximum CIR, and the minimum/maximum data rate are

time-dependent, but we dropped the time symbol (t) for simplicity In UMTS specifications

the power is updated on slot-by-slot basis The data rate is updated on a frame-by-framebasis To generalize the analysis, we use the same time symbol for power and rate

To increase the fairness, the users should achieve at least the minimum target CIR,

which corresponds to the minimum transmitted rate (e.g 15 kb/s in UMTS) The multirate

power control problem is defined as: given the target SINR vector δ = [δT

1, δT

2, , δT

the minimum requested data rate vector Rmin = [R1,min, R2,min, , RQ ,min], and without

users, find the optimum power vector P= [P1, P2, , PQ]and the optimum rate vector

R= [R1, R2, , RQ]that minimize the following cost function

N is the optimization time window, γ is a real-valued constant adaptation factor The notation

( )is used for transposed The error e i (t) has been defined according to the weighted metrics method

e i (t) = λ i ,1 |P i (t) − Pmin| + λ i ,2 | i (t) −  i ,min | + λ i ,3 | i (t) −  i ,max| (12.99)where 0≤ λ i ,1, λ i ,2, λ i ,3≤ 1 are real-valued, constant tradeoff factors,3

k=1λ i ,k = 1 The

advantages of joining the weighting metrics method with the least square formula of

Equation (12.97) are:

r The least squares method is well known and its derivation is straightforward.

r A general solution is obtained using Equation (12.97), minimizing over all users and for

time window N

The error function Equation (12.100) is the mathematical interpretation of the RRM

objectives given in (a)-(d) The first term of Equation (12.100) is to keep the transmitted power P i (t) as close as possible to Pmin, so we try to achieve objective (1) Objectives (2)and (3) will be achieved in the second part of the error function In this part, the transmittedpower is selected to obtain CIR very close to the minimum required CIR Achieving theminimum required QoS for every user maximizes the fairness in the cell The third term

in Equation (12.100) represents the objective (4), where the users will try to obtain themaximum allowed QoS if possible

By solving Equations (12.97) and (12.100) (using same procedure of MODPC algorithm)

for a one-dimensional (N = 1) case we obtain for i = 1, , Q:

P i (t+ 1) = λ i ,1 Pmin+ λ i ,2 i ,min + λ i ,3 i ,max

λ i ,1 P i (t) + (λ i ,2 + λ i ,3) i (t) P i (t) , t = 0, 1, , (12.100)and as before

R i (t+ 1) = Rs

δT

i

If the minimum solution places such demands on some users that they cannot be achieved,then dropping or the handoff process should be applied [84] The multirate power controlalgorithm given by Equations (12.100)–(12.103) has some interesting characteristics Bychanging the values of the tradeoff factorsλ i, different solutions with different meaningsare obtained For example, whenλ i ,1 = 1, λ i ,2 = 0, and λ i ,3= 0, it is clear that Equation

(12.100) will be reduced to a fixed level (no) power control and user i will send at minimum

power Forλ i ,1 = 0, λ i ,2 = 1 and λ i ,3= 0, Equation (12.100) becomes the distributed power

control (DPC) algorithm of Grandhi and Zander [82] In this case, the fairness is maximized.When λ i ,1 = 0, λ i ,2 = 0, and λ i ,3= 1, algorithm Equation (12.100) will maximize the

average transmitted rate (with using reasonable dropping algorithm for nonsupported users)

In this case one or a few users will be supported, so the outage probability will be high.From previous extreme conditions, one can make a tradeoff between these objectives to getthe best performance according to the required specifications The selection of the tradeoffvalues should be based on the communication link condition as well as the network andthe user requirements A wide range of different solutions can be obtained by changing thevalues of tradeoff factors The selection of one solution is a job for the decision maker

12.5.8 Multiobjective totally distributed power

and rate control (MOTDPRC)

In this section, we discuss a slight modification of the MODPRC algorithm, the totallydistributed algorithm The MODPRC algorithm, Equations (12.100)–(12.103), assumesthe availability of the actual CIR value In the existing and near-future cellular systems,

only an up–down command of the power is available at the MS The estimated CIR is usedwith the MOTDPRC algorithm The CIR (in dB) could be estimated as

˜

 i (t) dB = T

where ˜e i (t) is power control error, T

i (t) is the target CIR, and ˜  i (t) is the estimated CIR.

Using the estimated CIR in the MODPRC algorithm we obtain

P i (t)= λ i ,1 Pmin+ λ i ,2min + λ i ,3max

λ i ,1 P i (t − 1) + (λ i ,2 + λ i ,3) ˜ i (t) P i (t − 1), t = 0, 1, (12.104)

R i (t)= Rs

12.5.9 Throughput maximization/power minimization (MTMPC)

Another application of the MO optimization in the RRM can be achieved by modifying themaximum throughput power control algorithm The algorithm was based on maximizingthe throughput and ignoring the transmitted power levels In practice, reducing the trans-mitted power is very desirable In this section we will formulate the cost function with twoobjectives The first objective is the maximization of the total throughput as in Chawla andQiu [85] The second objective is to minimize the total transmitted power The approach istreating the total throughput maximization and the total power minimization simultaneouslyusing multiobjective optimization techniques

The problem is defined as follows: given the link gains of the users find the power vector which increases the total throughput as much as possible) and at the same time reduces the total transmitted power (as much as possible) The problem can be represented

mathematically as

where P= [P1, P2, , PQ]Tis the power vector, the objective functions

O1(P)=i Q=1 i , and O2(P)=i Q=1p i

and the admissible power set = {P|Pmin ≤ P i ≤ Pmax, i = 1, , Q} The minus sign

is used to minimize the second objective We will use the weighting method to solve themultiobjective optimization problem The idea of the weighting method is to associateeach objective function with a tradeoff factor (weighting coefficient) and maximizes (orminimizes) the weighted sum of the objectives [86] Applying the weighting method in ourproblem we obtain,

max

P {O(P)} s.t P ∈ , (12.108)where

is the multiobjective function, 1= [1, 1, , 1], and the tradeoff factors are real numbers,

To maximize the reward functions, Equation (12.111), we find the power vector P which

satisfies Equation (12.110) Since the obtained equations are nonlinear, it will be very

complicated to get an analytical solution An iterative solution for k = 1, , Q will be

formulated (we will drop the iteration argument t for simplicity)

Solving for P kleads to

reduces to Equation (12.78) Without power constraints Equation (12.78) is rewritten as

where P k ∈ [Pmin, Pmax], k = 1, , Q From Equation (12.119), the new transmitted

power is a scaled value of the transmitted power in the case of the maximum put algorithm To compare the two-algorithms, the numerical example from References

through-[85, 87] is used Consider the system with Q = 5 users and the path gain matrix, G, shown

The tradeoff factors have been set to{λ1, λ2} = {0.9999, 0.0001} In this case we penalize

power usage From Table 12.9, we can see that the summation of the SINR (dB) of the users[which is related to the total throughput as in Equation (12.78)] has not changed very much

in both schemes (only 0.04 %), but the power has been reduced by more than 98 % in thecase of MTMPC method Additional information on the topic can be found in References[88–90]

Table 12.9 Comparison of MTPC and MTMPC algorithms

12.6 DYNAMIC SPECTRA SHARING IN WIRELESS NETWORKS

In this section we present schemes for interference suppression in UWB-based WPAM tems when sharing the same band as other communications networks The scheme can beused to significantly improve the performance of UWB systems in the presence of interfer-ence from mobile communication systems such as GSM and WCDMA It is also effective

sys-in the presence of WLAN systems which are nowadays based on OFDMA technology or sys-inmilitary communications where the interference is generated by intentional jamming Thesection also discuss the effectiveness of the scheme to suppress MC CDMA, which is acandidate technology for 4G mobile communications In order to demonstrate the relevancy

of these results we first provide a systematic review of the existing work in this field andthen present specific scheme The results show that significant suppression gain up to 40 dBcan be achieved in the presence of OFDM, WCDMA and MC CDMA, enabling coexistence

of different networks in the same frequency bandwidth The effectiveness decreases if thenumber of subcarriers is increased

The online source [91] gives historical perspective to UWB technologies It lists downthe early UWB references and patents from the 1960s and 1970s In [92] a comprehensiveoverview of UWB wireless systems is given It discusses the FCC allocation of 7.5 GHz(3.1–10.6 GHz) unlicensed band for the UWB devices Potential UWB modulation schemes,multiple access issues, single vs multiband implementation and link budgets are also dis-cussed Paper [93] is a very frequently referenced one giving a brief introduction to thebasics of impulse radio systems It describes the characteristics of impulse radio and givesanalytical estimates of the multiaccess capability under idealistic channel conditions

12.6.1 Channel capacity

Some new channel capacity results for M-ary pulse position modulation (M-PPM) time

hopping UWB systems are presented in [94] It is demonstrated that the previous resultsbased on the ‘pure PPM model’ have overestimated the real UWB capacity The proposedmodel is extended with the correlator and soft decision decoding The capacity is evaluated

in the single-user case and with asynchronous multiple user interference (MUI) when the

inputs are equiprobable It is found that larger M leads to increased capacity only at the high bit SNR region Furthermore, optimal time offset values for each M are independent of the

bit SNR The MUI influence is detrimental for the capacity, especially at high bit SNRs

12.6.2 Channel models

Paper [95] focuses on the UWB indoor channel modeling issues The measurement data

is collected from an extensive campaign in a typical modern office building with a 2 nsdelay resolution The model is formulated as a stochastic tapped delay line (STDL) Theenergy statistics due to small-scale effects seem to follow a Gamma distribution for allbins Large-scale parameters can be modeled as stochastic parameters that can change, e.g.,from room to room UWB propagation channels are also discussed in [96] Based on themodified CLEAN algorithm, estimates of time-of-arrival, angle-of-arrival, and waveformshape are derived Key parameters of the model are intercluster decay rate, intraclusterdecay rate, cluster arrival rate, ray arrival rate and standard deviation of the relative azimutharrival angles Intercluster signal decay rate is generally determined by the architecture

of the building Intracluster decay rate, on the other hand, depends on the objects close

to the receive antenna (e.g furniture) Relative azimuth arrival angles were best fit to aLaplacian probability density function Saleh and Valenzuela [97] present a model that hasbecome a frequently referenced and adopted source in indoor multipath propagation channelmodeling They propose a statistical indoor radio channel model that (1) has flexibility toreasonable fit with the measured data, (2) is simple enough to be used in simulation andanalysis, and (3) can be extended by adjusting parameters to represent various buildings Inthe developed statistical model the rays of the received signal arrive in clusters

The ray amplitudes are independent Rayleigh random variables with exponentially caying variances with respect to the cluster delay and the ray delay The clusters and therays within the cluster form Poisson arrival processes with different, fixed rates Paper [98]characterizes measurement-based UWB wireless indoor channels from the communicationstheoretic viewpoint The bandwidth of the signal used in the measurement is over 1 GHz,resulting in the less than 1 ns time resolution Robustness of the UWB signal to multipathfading is quantitatively evaluated through histograms and cumulative distributions Tworake structures are introduced: the all rake serves as the best-case (benchmark) receiverand the maximum-energy-capture selective rake is a realistic sub-optimal approach Mul-tipath components of the measured waveforms are detected using a maximum-likelihoodestimator based on a separable specular multipath channel model

de-12.6.3 Diversity reception

Performance of PPM and on–off keying (OOK) binary block-coded modulation formatsusing a maximal ratio combining rake receiver is studied analytically in [99] The trade-offbetween receiver complexity and performance is examined Several suboptimal receivers

in indoor multipath AWGN channels have been employed Results indicate the robustperformance may require an increase in rake complexity This implies allocation of morerake fingers and tracking of the strongest multipaths to help in the selection combining Rakeperformance for a pulse-based high data rate UWB system in an Intel Labs indoor channelmodel is addressed in [100] It is noted that, at low input SNR values (0–10 dB) and small

number of rake fingers, it is more beneficial to add rake taps for energy capture rather thanfor intersymbol interference (ISI) mitigation In the presence of channel estimation errors,equal gain combining can be more robust than maximal ratio combining, and thereforeyield better performance In order to quantify the trade-off between rake receiver energycapture and diversity order [101] presents partly quasi-analytical and partly experimentalanalysis suited to dense multipath propagation environments Numerical results show that

a diversity level of less than 50 is adequate in typical indoor office conditions

12.6.4 Performance evaluation

In [102] a method to evaluate the BER performance of time hopping (TH) PPM in thepresence of multiuser interference and AWGN channel is proposed Gaussian quadraturerules are used in this approach Paper [103] concentrates on the signal design for binaryUWB communications in dense multipath channels The aim is to find signals with gooddistance properties leading to good BER performance that both depend on the time shift

monocyles is studied in [104] Channel conditions vary among ideal single user AWGN,nonideal synchronous, multipath fading and multiple access interference It is shown thatthe pulse shape has a notable impact on the correlation receiver performance The effectscan be seen in the autocorrelation function, especially in the mainlobe The autocorrelation

is highly related to the SNR gain of the output and to interference resistance properties.Special characteristics of the Gaussian monocycles include: (1) higher order derivativeshave higher SNR gain in single user and asynchronous multiple access channel but are lessrobust to interference than lower order derivatives; (2) narrower pulses have higher SNRgain in asynchronous multiple access channel at the cost of inferior interference resistanceability Exact bit error rate performance of TH-PPM UWB systems in the presence ofmultiple access interference (MAI) is analyzed and simulated in [105] Furthermore, it isshown that, with a moderate number of MAI sources, the standard Gaussian approximationbecomes inaccurate at high SNRs

12.6.5 Multiple access techniques and user capacity

The main principles for multiaccess in UWB systems are discussed in [106] A functionalmedium access, radio link and radio resource control architecture is proposed and openissues for future activities are addressed Numerical throughput and delay performanceresults in radio resource sharing are shown Uncoded and coded performance analysis forTH-UWB systems is covered in [107] A practical low-rate error correcting coding scheme

is presented that requires no bandwidth expansion Gaussian assumption for multiuser terference is shown to be invalid for high uncoded data rates The user capacity is shown

access capacity analysis Performance is analyzed in free-space propagation conditions.The number of supported users is dependent on the given bit error rate, SNR, transmis-sion rate and modulation alphabet size Performance and receiver complexity trade-off isdiscussed Upper bounds are derived for the combinations of user capacity and total trans-mission rate According to the numerical examples it is possible to achieve a system with

high capacity and data rate, low bit error probability, and yet only moderate receiver plexity Reference [109] is one of the first public and widely cited papers outlining thepotential of time hopping impulse radio multiple access communications It describes thebasic building blocks of the impulse transmitter and receiver and their mathematical for-mulations It also shows an example for the bit error vs user capacity estimate at variabledata rates Finally, some drawbacks of the high time-resolution impulse radio systems arementioned: (1) the need for up to thousands of rake fingers in the multipath receiver; and(2) complex initial clock acquisition In [110] a comprehensive overall description of thetime-hopping UWB system physical layer issues is given Achievable transmission ratesand multiple access capacities are estimated for analog and digital modulation formats.Numerical results indicate that the digital implementation has the potential for nearly oneorder of magnitude higher user densities than the analog one.

com-12.6.6 Multiuser detection

Reference [111] is focused on the multiuser detection (MUD) possibilities for directsequence UWB systems It is demonstrated that the adaptive minimum mean squared er-ror (MMSE) MUD receiver outperforms the rake receiver both in energy capture and ininterference rejection sense Studied interference sources are narrowband IEEE 802.11a in-terference and wideband multiuser UWB interference Ideally, MMSE receiver can achieveAWGN bit error rate within a 1–2 dB margin even in dense multipath channels In heav-ily loaded conditions the penalty of 6 dB is experienced, but at the same time the rakereceivers suffer from unbearable error floors Iterative partial parallel multiuser interfer-ence cancellation (PIC) is applied to the UWB multiuser system in [112] Matched filter,maximum-likelihood, and linear minimum mean squared error receivers are also used inthe performance comparison In this paper, multiuser detection is combined with error con-trol coding The UWB system includes only one pulse per symbol and AWGN channel isassumed Numerical results show that it is possible to attain the coded single user BERbound for eight to 15 users in a heavily loaded system without any processing gain Asthe number of users increases and the bandwidth to pulse repetition frequency decreases,MAI is expected to adversely affect system capacity and performance As a consequence, aframework for the design of multiuser detectors for UWB multiple access communicationssystems is presented in [113] An optimum multiuser detector is also proposed

12.6.7 Interference and coexistence

Coexistence of UWB system with some other radio systems is studied in [114] This meansthe evaluation of interference caused by the UWB system to the other radio systems and viceversa The coexisting radio concepts are GSM900, UMTS/WCDMA and GPS Several shortGaussian-based UWB pulses are employed According to the numerical results, convenientselection of pulse waveform and width leads to interference resistance up to a certainlimit The pulse shape is in interaction with the data rate High-pass filtered waveforms arepreferred in the case of short UWB pulses, whereas generic Gaussian ones are favorable iflong pulses are utilized Interference caused by narrowband systems is the most detrimental

to UWB if it is located at the UWB system’s nominal center frequency In the GPS bandthe DS based UWB system interfered less than the time hopping system

12.6.8 Channel estimation/imperfections

Channel estimation for time hopping UWB communications is dealt with in [115].Multipath propagation and MAI are taken into consideration Maximum-likelihood esti-mation is applied in data-aided and nondata-aided scenarios Numerical results show thatthe performance is reasonable if the number of simultaneous users is below 20 The impact

of the timing jitter and tracking to the impulse radio system performance is investigated in[116] Binary and 4-ary modulations are used According to these studies both modulationssuffer from the jitter, however 4-ary is better Timing jitter is also discussed in [117] Staticand Rayleigh fading channels are assumed Orthogonal PPM, optimum PPM, OOK andBPSK modulations are compared in performance evaluation Similar performance degra-dation is noted for BPSK and PPM schemes, while OOK is more susceptible to large jitter.The probability density function of timing jitter due to rake finger estimation is simulated.The results depend on the pulse shape and SNR Worst case distribution is shown to provide

an upper bound for BER performance The above survey of issues in UWB communicationsindicates a need for special attention to interference avoidance or interference suppressiondue to extremely wide signal bandwidth and the possibility of interference with other sys-tems operating in the same bandwidth One way to deal with the problem is to designthe pulse shape in such a way that the signal has no significant spectral component in theoccupied frequency bands Pulse shapes respecting the FCC spectral mask were proposed

in References [118–120] The drawback of such a solution is the need for over samplingand lack of flexibility in the case that the interfering signal is not stationary like in militaryapplications

Another approach is to use adaptive interference suppression like the schemes

sum-marized in the previous sub-section entitled interference and co-existence The solution

discussed in this section belongs to the latter category We will demonstrate the advantages

of this approach with a number of numerical results The solution is adaptive and can beimplemented with no over-sampling, unlike other schemes

12.6.9 Signal and interference model

In general the signal transmitted by the desired user is modeled as:

i

b[t − i N Tf− (1 − a i)] cos ωc t (12.120)where

The signal can also be transmitted in the baseband with no carrier In Equation (12.121),

g(t) represents the basic pulse shape (monocycle pulse) and Tf represents frame duration

during which there is only one pulse Tcseconds wide The sequence h(n) is the user’s

The parameter Tcis the duration of an addressable time bin within a frame In other words,

the right-hand side of Equation (12.121) consists of a block of N time-hopped monocycles.

a i represents information bits (0,1) Equation (12.121) says that, if a i were all zero, the

the time shift impressed by a unit data symbol on the monocycles of a block It is clearthat the choice of affects the detection process and can be exploited to optimize system performance In summary, the transmitted signal consists of a sequence of b(t)-shaped

position-modulated blocks

The code sequence restarts at every data symbol This ‘short-code’ assumption is madehere for the sake of simplicity and is in keeping with some trends in the design of third-generation CDMA cellular systems Longer codes are conceivable and perhaps more at-tractive but lead to more complex channel estimation schemes The OFDM interference,generated for example by a WLAN user, is modeled as

where N is the number of channels, d i is the FDM interference information bits, fc+ fc

is the first channel carrier frequency, J i is the OFDM interference amplitudes,ϕ i is the

channel phase at the receiver input and T jis the bit interval

The MC CDMA interference can be modeled as

received over a channel with Lcpaths, the composite waveform at the output of the receiverantenna may be written as:

B filter Finger 1

Figure 12.22 Receiver block diagram

The detection variable in the lth rake receiver finger is:

The weight of the I branch in the lth finger is:

T I l (i ) = D AI l (i ) + DB I l (i ) (12.135)

and the weight of the Q branch in the lth finger is:

T Q l (i ) = D AQ l (i ) + DB Q l (i ) (12.136)

12.6.11 Interference rejection circuit model

Interference rejection at UWB radio system may be performed by a transversal filteremploying LMS algorithm Basically, in the first step, the interfering signal is estimated inthe presence of the UWB signal which is at that stage considered as an additional noise

The estimated interference ˆj is subtracted from the overall input signal r , creating the input signal r= r − ˆj = s + n + j − ˆj = s + n + j to the standard UWB receiver In order

to predict the interference signal, sampling is performed at frame rate, and the adaptation

of filter weights using LMS algorithm is performed at bit rate It is already known that thechanges of the symbol in the interfering signal will disrupt the estimation process Curve 3,

in Figure 12.23, shows the detection variable at the output of the transversal filter when there

-6 -4 -2 0 2 4 6

with interference rejection circuit; 3, interference with J : S= 40 dB, with

 = 5 ns, vbJ = 100 Msymbol/s, vbTH= 5 Mbt/s, Tframe= 10 ns, fc=

800 MHz

Frame rate processing

Bit rate processing

Figure 12.24 Interference rejection circuit block diagram TH j (t)= time hopping sequence

[TH j (t) = b(t – iNTf− j), j = 1 for filter A, j = 0 for filter B]; SH =

sample and hold circuit; BPF(F), BPF(B)= backward prediction filter; FPF(F),FPF(B)= forward prediction filter; SEL LOG = selection logic; Ai = am-

plifier, i= 1, 2; and C i = comparator, i = 1, 2.

is PSK interference (with interference to signal ratio of 40 dB) at the input of the receivertogether with the useful UWB signal and Gaussian noise The presence of an impulse inter-ference may be seen in the figure The appearance of the impulse interference is very similar

to the one at DSSS system using transversal filter for interference rejection [121] A ‘U’structure, that successfully rejects this impulse interference, was proposed in the mentionedreference Similarly, in this section we discuss a structure, shown in Figure 12.24, for theinterference rejection, which is based on the ‘U’ structure and has been modified to matchthe UWB radio environment

Curve 2 in Figure 12.23 shows the signal at the output of the interference rejectionstructure in the case when there is PSK interference with interference-to-signal ratio of

40 dB at the input of the receiver It can be noted that the interference is rejected and thedetection variable is very similar to the one when there is no interference (curve 1).More details on operating principle of the interference rejection circuit is shown inFigures 12.24 and 12.25 We consider the case of a bit interval having 20 frames with five

Figure 12.25 Signal processing.

samples per frame (M= 2) Central samples, within each frame, carry the same information

about the useful signal and the interference symbols, and, since each sample belongs to adifferent frame, all those samples originate from different instances of time Similarly,samples from different frames equally distant from the central sample carry the correlatedinterference signal Therefore, an equivalent signal may be formed in the following way:

Mth equivalent signal sample is the sum of Mth samples from each frame Adaptation

of filter weights using LMS algorithm and interference prediction is performed using theequivalent signal samples

As already mentioned, the changes of the symbol in the interfering signal will disruptthe estimation process so that a forward and backward prediction are used simultaneously.When the symbol change occurs, the filter with less disruption (smaller error at its output)

is used to deliver the estimates This process is described in the following in more detail.For additional insight into the problem the reader is also referred to Cox and Reudink [31].Possible moments of the transition to happen are shown in Figure 12.25 and are denotedwith 1◦, 2◦, 3◦ and 4◦ Therefore, at frame rate BPF(F) (backward prediction filter) andFPF(F)(forward prediction filter) filters are operating with weights being forwarded fromthe LMS algorithm adapted by the equivalent signal BPF(F) and FPF(F) filters have the

same weights during the useful signal bit interval, i.e within all the frames belonging to thesame bit interval So the interference is predicted using the same weights computed usingthe equivalent signal If, in cases 1◦, and 2◦, we discard samples belonging to the BPF(F)filter we will also discard the interference transition influence on the prediction For cases 3◦and 4◦, samples that belong to the FPF(F)filter should be discarded This sample discarding

is performed in the selector S at the outputs of BPF(F)and FPF(F)filters, based on the errorsignal

Therefore, if there is no interference signal transition during the sampling within oneframe, the equivalent signal will be formed using all the samples from that frame Also,the same equivalent central sample (SX0(B)= SX0(F)) is passed to both BPF(B)and FPF(B)LMS algorithms On the other hand, if there is interference signal transition during thesampling within one frame, the equivalent signal will be formed using samples from FPF(F)(cases 1◦and 2◦) or BPF(F) (cases 3◦ and 4◦), and there will be two different equivalentcentral samples

For the described Interference rejection circuit we have: the first part of the interference rejection circuit processes data at frame level, and at each frame the following input signal

processing is performed At filter A, the signal is sampled very close in time to the useful

After that, variables C1 and C2, which the operation of one side of filter X (A and B) is

based on, are determined:

where A i (i = 1, 2) are constants (for A i→ ∞, the selector selects all the samples and the

structure operates as a traditional LMS algorithm) These gains are introduced because ofthe decrease of noise influence on the irregular selections

The second part of the interference rejection circuit operates at bit interval level

Tb = N Tf, and for each ith bit we have the following signals:

The adaptation algorithm is defined as:

N iis the bit ensemble size (measured in number of information bits) used for averaging the

result and Nais the number of ensemble members

Estimated signal-to-noise ratio per bit is:

Figure 12.26 presents the results for BER as a function of signal to noise ratio SNR,

in the presence of different PSK/QAM type interfering signals Additional parameters of

g c

b, b1

a

Pe

SNR (dB)Figure 12.26 Error probability as a function of signal-to-noise ratio Error probability based

on Monte-Carlo simulation: a, no interference, without interference rejection filter; b, no interference, with interference rejection filter; c, PSK interference, J: S = 40 dB, with interference rejection filter; d, QPSK interference, J: S =

40 dB, with interference rejection filter; e, 16QAM interference, J: S= 40

dB, with interference rejection filter; f, 64QAM interference, J: S= 40 dB,

with interference rejection filter; g, 256QAM interference, J: S= 40 dB, with

interference rejection filter; Error probability based on estimated detection

variable signal to noise ratio: b1, the same parameters as b; d1, the same

parameters as d; f1, the same parameters as f; fc = 800 MHz, M = 4,  =

5 ns,vbJ = 100 Msymbol/s, vbTH= 5 Mbt/s, Tframe = 10 ns

fc = 800 MHz and Tframe= 10 ns One can see: (1) fair agreement of simulation and

numerical results; (2) the performance results are close to no interference case, althoughinterference with the level of 40 dB above the UWB signal is present; (3) there is also a slightdegradation of the performance when the interfering signal constellation size is increased

Figure 12.27 presents the results for BER as a function of interference to signal ratio J:S,

in the presence of different PSK/QAM-type interfering signals Additional parameters of

the signals are: filter length M = 4, SNR=7 dB,  = 5 ns, v bJ = 100 Msymbol/s, vbTH= 5

Mbt/s,fc = 800 MHz and Tframe = 10 ns One can see that, when J:S becomes larger than

zero (5 dB), the BER increases rapidly if there is now interference suppression (curves A).The performance is very similar if a standard LMS algorithm is used (curves C) On the otherhand the U-type filter is performing significantly better (curves B) There is again a slight

J:S (dB)

Figure 12.27 Error probability as a function of interference-to-signal ratio A, without

in-terference rejection; B, with inin-terference rejection circuit; C, with classical LMS interference rejection filter; a, PSK interference; b, QPSK interference;

c, 16QAM interference; d, 64QAM interference; e, 256QAM interference;

fc = 800 MHz, M = 4, SNR = 7 dB,  = 5 ns; v bJ = 100 Msymbol/s,

vbTH = 5 Mbt/s, Tframe = 10 ns

degradation of the performance when the interfering signal constellation size is increased

Figure 12.28 presents the results for BER as a function of interference symbol duration T j /Tc

in the presence of different PSK/QAM-type interfering signals Additional parameters of the

and Tframe = 10 ns One can see that BER decreases when T j /Tcincreases There is again

a slight degradation of the performance when the interfering signal constellation size isincreased Figure 12.29 presents the results for BER as a function of interference symbol

duration T j /Tc and the number of subcarriers N in the presence of OFDM-type interfering

can see that BER decreases when T j /Tc increases up to T j /Tc≈ 200 Beyond that point

there is no significant reduction in BER if T j /Tcis further increased There is a significantdegradation of the performance when the number of subcarriers in the OFDM signal isincreased

5 10 15 20 25 30 35 40 0.01

d e

c b

a

Pe

T j /Tc

Figure 12.28 Error probability as a function of interference bit duration a, PSK

ference; b, QPSK interference; c, 16QAM interference; d, 64QAM

7 dB; M = 4;  = 5 ns; vbTH= 5 Mbt/s; Tframe = 10 ns

200 400 600 800 1000 0.01

0.1

4

64 32 16 8

Figure 12.29 Error probability as a function of OFDM interference bit duration and the

454

200 400 600 800 1000 0.01

Figure 12.30 Error probability as a function of OFDM interference bit duration A,

in-terference rejection circuit; B, classical LMS inin-terference rejection filter; a, OFDM/PSK interference; b, OFDM/QPSK interference; c, OFDM/16QAM interference; d, OFDM/ 64QAM interference; e, OFDM/256QAM interfer-

ns;vbTH = 5 Mbt/s, Tframe= 10 ns

Figure 12.30 presents the results for BER as a function of interference symbol duration

= 10 ns , fc= 800 MHz and N = 16 One can see again that BER decreases when T j /Tc

increases There is again a slight degradation of the performance when the interferingsignal constellation size is increased Once again, the U-type filter performs much betterthan the classical LMS algorithm Figure 12.31 presents the results for BER as a function of

interference symbol duration T j /Tcin the presence of MC CDMA-type interfering signals

for different number of subcarriers N and number of users K Additional parameters of the

performance are improved if the number of subcarriers is decreased

1000 2000 3000 4000 5000 6000 7000 8000 0.01

0.1

e d

b a

c

Pe

T j /T c Figure 12.31 Error probability as a function of MC CDMA interference bit duration a,

In this section we presented a U-type estimation filter based scheme for interferencesuppression in UWB systems and discussed its performance It was shown that the schemecan be used to significantly improve the performance of UWB systems in the presence ofinterference from mobile communication systems such as GSM and WCDMA It is alsoeffective in the presence of WLAN systems which are based on OFDMA technology or inmilitary communications where the interference is generated by intentional jamming Thesection also discusses the effectiveness of the scheme to suppress MC CDMA, which is

a candidate technology for 4G mobile communications The results show that significantsuppression gain up to 40 dB can be achieved in the presence of OFDM, WCDMA and MCCDMA The effectiveness decreases if the size of the number of subcarriers is increased

[1] S.G Glisic and P Pirinen, Co-Channel Interference in Digital Cellular TDMA Networks, John Wiley Encyclopedia of Telecommunications, ed J Proakis John

Wiley & Sons Ltd: Chichester, 2003

[2] J Zander, Asymptotic bounds on the performance of a class of dynamic channel

assignment algorithms, IEEE JSAC, vol 11, 1993, pp 926–933.

[3] J C.-I Chuang, Performance issues and algorithms for dynamic channel assignment,

IEEE JSAC, vol 11, 1993, p 6.

[4] T.J Kahwa and N Georganas, A hybrid channel assignment scheme in large scale

cellular-structured mobile communication systems, IEEE Trans Commun., vol COM

26, 1978, pp 432–438

[5] J Sin and N Georganas, A simulation study of a hybrid channel assignment scheme

for cellular land mobile radio systems with Erlang C service, IEEE Trans on mun., vol COM-9, 1981, pp 143–147.

Com-[6] D Cox and D.O Reudink, Increasing channel occupancy in large scale mobile radio

systems: dynamic channel reassignment, IEEE Trans Vehicular Technol., vol.VT-22,

1973, pp 218–222

[7] D Cox and D.O Reudink, A comparison of some non-uniform spatial demand

pro-files on mobile radio system performance, IEEE Trans Commun., vol COM 20,

1972, pp 190–195

[8] D.C Cox and D.O Reudink, Dynamic channel assignment in two dimensional

large-scale mobile radio systems, Bell Syst Tech J., vol 51, 1972, pp 1611–1628.

[9] M Zhang and T.-S Yum, The non-uniform compact pattern allocation algorithm for

cellular mobile systems, IEEE Trans Vehicular Technol vol VT-40, 1991, pp 387–

391

[10] S.-H Oh and D.W Tcha, Prioritized channel assignment in a cellular radio network,

IEEE Trans Commun., vol 40, 1992, pp 1259–1269.

[11] T Anderson, A simulation study of sole dynamic channel assignment algorthms in

high capacity mobile telecommunications system, IEEE Trans Vehicular Technol.,

vol VT-22, 1973, p 210

[12] J.S Engel and M Peritsky, Statistically optimum dynamic server assignment in

systems with interfering servers, IEEE Trans Vehicular Technol., vol VT-22, 1973,

pp 203–209

[13] M Zhang, Comparisons of channel assignment strategies in cellular mobile telephone

systems, IEEE Trans Vehicular Technol., vol VT38, 1989, pp 211–215.

[14] R Singh, S.M Elnoubi and C Gupta, A new frequency channel assignment

algo-rithm in high capacity mobile communications systems, IEEE Trans Vehicular Tech.,

vol VT-31, 1982

[15] P John, An insight into dynamic channel assignment in cellular mobile

communica-tion systems, Eur J Opnl Res., vol 74, 1994, pp 70–77.

[16] S Tekinay and B Jabbari, Handover and channel assignment in mobile cellular

networks, IEEE Commun Mag vol 29, 1991, pp 42–46.

[17] T.-S P Yum and W.-S Wong, Hot spot traffic relief in cellular systems, IEEE JSAC,

vol 11, 1993, pp 934–940

[18] S Kuek, Ordered dynamic channel assignment scheme with reassignment in highway

microcell, IEEE Trans Vehicular Technol., vol 41, 1992, pp 271–277.

[19] K Okada and F Kubota, On dynamic channel assignment in cellular mobile radio

systems, Proc IEEE Int Symp Circuits and Systems, vol 2, 1991, pp 938–941.

[20] K Okada and F Kubota, On dynamic channel assignment strategies in cellular mobile

radio systems, IEICE Trans Fundamentals, vol 75, 1992, pp 1634–1641.

[21] D Cox and D Reudink, A comparison of some channel assignment strategies in

large mobile communication systems, IEEE Trans Commun., vol 20, 1972, pp 190–

195

[22] D Cox and D Reudink, Dynamic channel assignment in high capacity mobile

com-munications systems, Bell Syst Tech J., vol 50, 1971, pp 1833–1857.

[23] A Gamst, Some lower bounds for a class of frequency assignment problems, IEEE Trans Vehicular Technol., vol 35, 1986.

[24] K Okada and F Kubota, A proposal of a dynamic channel assignment strategy with

information of moving directions, IEICE Trans Fund., vol E75-a, 1992, pp 1667–

1673

[25] V Akaiwa and H Andoh, Channel segregation-A self organized dynamic allocation

method: application to TDMA/FDMA microcellular system, JSAC, vol 11, 1993,

pp 949–954

[26] V Prabhl and S.S Rappaport Approximate analysis of dynamic channel

assign-ment in large systems with cellular structure, IEEE Trans Commun., vol 22, 1974,

pp 1715–1720

[27] J.B Punt and D Sparreboom, Mathematical models for the analysis of dynamic

chan-nel selection or indoor mobile wireless communications systems, PIMRC, vol E6.5,

1994, pp 1081–1085

[28] D Hong and S Rappaport, Traffic modelling and performance analysis for cellularmobile radio telephone systems with prioritized and nonprioritized handoff proce-

dures, IEEE Trans Vehicular Technol., vol VT 35, 1986, pp 77–92.

[29] Y Furuya and V Akaiwa, Channel segregation, A distributed channel allocation

scheme for mobile communication systems, IEICE Trans., vol 74, 1991, pp 1531–

1537

[30] J Vucetic, A hardware implementation of channel allocation algorithm based on

a space-bandwidth model of a cellular network, IEEE Trans Vehicular Technol.,

vol 42, 1993, pp 444–455

[31] D Cox and D Reudink, Increasing channel occupancy in large scale mobile

ra-dio systems: dynamic channel reassignment, IEEE Trans Commun., vol 21, 1973,

pp 1302–1306

[32] W Yue, Analytical methods to calculate the performance of a cellular mobile

ra-dio communication system with hybrid channel assignment, IEEE Trans Vehicular Technol., vol VT-40, 1991, pp 453–459.

[33] J Tajime and K Imamura, A strategy for flexible channel assignment in mobile

communication systems, IEEE Trans Vehicular Technol., vol VT-37, 1988, pp 92–

103

[34] J Zande and J Frodigh, Capacity allocation and channel assignment in cellular radio

systems using reuse partitioning, Electron Lett., vol 28, 1991.

[35] S W Halpern, Reuse partitioning in cellular systems, IEEE Trans Vehicular Technol.,

1983, pp 322–327

[36] T Kanai, Autonomous reuse partitioning in cellular systems, IEEE VTC, 1992

pp 782–785