44

CHAPTER 3

CHANNEL EFFICIENCY ENHANCEMENT IN WCDMA

SYSTEMS USING SOFT HANDOFF

3.1 INTRODUCTION

Mobile radio communication has become a rapidly growing market since

the GSM standard has been established. Meanwhile, 3G mobile radio systems like

UMTS and CDMA 2000 have been standardized and the 4G is currently under

investigation. During this development, WCDMA has become a widely accepted

multiple access technique. In most cases, it is implemented as Direct-Sequence

CDMA (DS-CDMA) in single-carrier systems.

An important feature of cellular mobile communication is handoff. It is

defined as the transfer of a user's connection from one radio channel to another

(can be the same or different cell). As mobile moves towards the boundary of its

serving cell, the movement causes dynamic changes in the interference levels and

the link quality. This may cause the mobile to transfer communication to or to

migrate to a different BS. This change of serving BS is called a handoff. Hard

Handoff, also known as a `break-before-make' handoff, is the category of handoff

procedures in which the mobile switches to a new radio link after breaking

connection with the old radio link [13]. At any time, the active set (set of BSs with

which the user is in communication) will have only one BS. SHO on the other hand,

is the handoff procedure in which a mobile has connection with more than one radio

link simultaneously during handoff.

Once the signal from a single radio link is considerably stronger than the

others, a decision will be made to communicate with that one only. The addition and

45

removal of BSs into and from the active set is dependent on parameters such as the

add threshold (TADD), drop threshold and drop timer. A BS is added to the active set

when its pilot signal strength exceeds the add threshold. A BS is removed from the

active set when its pilot signal strength drops below the drop threshold and

stays below it for the time specified by the drop timer. The process is illustrated in

Figure 3.1.

Figure 3.1 Soft handoff procedure

3.2 CHANNEL CONVERTIBLE SET (CCS)

3.2.1 Cell Geometry of SHO

Accurate geometry of SHO in WCDMA cellular system is hard to depict

due to various factors, such as irregular cell boundaries, traffic conditions and the

movement of mobiles. To simplify the problem, the following reasonable

assumptions are made.

1) The cellular system includes a number of cells of identical size and shape

and the coverage area of each cell can be approximated by a circle.

46

2) Mobiles initiating the calls are uniformly distributed throughout each cell

and one mobile unit may carry at most one call at a time.

3) The cells in the system are symmetrically located and well distributed.

Each cell is surrounded by six other cells.

4) An MS in handoff area occupies at most two channels in its active set,

i.e., there are at most two different sources in diversity reception.

Figure 3.2 Cellular structure of Soft handoff

Figure 3.2 illustrates an example of regions and boundaries based on the

assumption of a circular cell area. With regard to signal strength [102, 103], a cell

can be divided into two areas, namely 1) the normal area and 2) handoff area. In the

SHO area, represented by the intersection of target cell and neighbourhood cell,

each MS holds two channels for transmission in diversity. The ratio ah of the

handoff area to the entire cell area is defined as

hthe area of the handoff regiona

the area of the cell (3.1)

The handoff area is further divided into two regions as shown in

Figure 3.3; target controlling (TC) region and neighbour controlling (NC) region.

In the TC region, the BS of a target cell has stronger power than that of a

neighbouring cell in active sets of MSs. In the NC region, the BS of a neighbouring

47

cell has stronger power than that of a target cell in active sets of MSs. Basically,

since selection diversity is used for uplink interference in WCDMA systems; the BS

that has a higher receiving power in the active set of a call plays an important role in

demodulating the received signal.

Figure 3.3 Cellular system model of Soft handoff

3.2.2 Relative Mobility Estimation

It is assumed that the received pilot strength from a BS decreases when

the MS moves away from the BS and increases when the MS moves towards the BS.

An MS in the handoff area can detect pilot strength from the serving BS and current

time t broadcast by the sync channel which is the forward link channel used to

transmit some system parameters. A mobile uses this information for time

synchronization, which is crucial for the mobile to establish a forward traffic

channel with the BS. Let ps(t,i) be the received pilot strength from the serving BS,

which is measured at time t by MSi and cr_ps(t,i) be the rate of change of ps(t,i)

given as

( , ) ( , )_ ( , ) ps t t i ps t icr ps t it

(3.2)

where, ∆t is the time period of information update in the cellular system.

TC-region NC-region

Cell A MS2

Cell B

MS1

48

With the pilot strength and the rate of changing, the relative mobility of

calls in the handoff area, such as relative position, moving direction and velocity can

be estimated [101, 104]. It is reasonable to assume that the BS with stronger pilot

strength ps(t,i) in the active set of the MSi in the process of handoff should be nearer

to the MS than the BS with weaker pilot strength. In addition, a handoff call

must be moving toward the BS if the rate of change cr_ps(t,i) detected by the BS is

positive. The bigger the value of cr_ps(t,i), the higher the velocity of the MS.

If |cr_ps (t,i)|<ε, where ε is a suitably chosen small number, the MS is considered to

be stationary. The stationary calls in the handoff area request multiple channels,

even though the call is not actually approaching the neighbouring BS from the target

BS.

Therefore, soft handoff can be performed in terms of both the received

signal strength and measured relative mobility of calls [99]. The defined cellular

areas can be identified by measuring ps(t,i) and cr_ps(t,i). The coverage area of a

cell is determined by checking if the pilot strength (ps(t,i)) of the call in the area is

greater than TADD. The SHO area of two cells is determined by checking if more than

two pilot strengths from both the target BS and the neighbouring BS at time t are

greater than TADD. Therefore, the normal area of a target cell is determined by

checking the ps(t,i) from the target BS is greater than TADD and the pilot strengths

from all neighbouring BSs are less than TDROP. Furthermore, in the handoff area the

TC and NC regions can be distinguished by comparing the measured pilot strength.

Let psT(t, i) be the strength of the pilot received from the target BS and let psN(t,i) be

the strength of the pilot received from the neighbouring BS. If psT(t, i)>psN(t,i), the

MS must be in the TC region, in which calls are mainly controlled by the target BS.

If psT(t,i)<psN(t,i), the MS must be in the NC region, in which calls are controlled by

the neighbouring BS.

It should be noted that relative mobility estimation is essentially

different from mobility management which is utilized in current wireless cellular

systems [100]. The main purpose of mobility management is to support the

registration and deregistration of an MS by using a Direct-Transfer Application Part

49

(DTAP) message whereas relative mobility estimation is concerned with the relative

position, speed and moving direction of an MS in a SHO area needed for

constructing the CCS in terms of the measured signal strength [35].

3.2.3 Pseudo Handoff Calls

The set that contains all handoff calls in the NC-region of the target cell,

such that they have channels from both the target BS and neighbouring BS and they

are staying stationary or moving away from the target cell, is defined as the CCS of

the target cell. According to the preceding definition, a CCS includes three types of

SHO calls:

1) The first are new calls that originated in the NC-region of the SHO area and

are moving away from the target cell. These calls request SHO to the

neighbouring BS immediately after their new calls are accepted by the

serving BS. The number of such calls is about 25% of the total number of

new calls in the handoff area [102].

2) The second are handoff calls that stay stationary while talking. In real urban

WCDMA systems, the number of stationary MS calls is about 40%–50% of

the total number of the MS calls.

3) The third calls that move from the TC-region to the NC-region of the target

cell while talking and continue to move towards a neighbouring cell. In this

case, the BS of the neighbouring cell has been changed as the controlling BS

of the call during the handoff process.

The calls in the first two cases are defined as pseudo handoff calls

because those calls do not really carry out handoff. In each BS, a CCS is set up to

identify pseudo handoff calls and serve for incoming handoff calls.

50

3.2.4 Construction and Updating of CCS

The CCS is constructed and managed as shown in Figure 3.4(a). For each

accepted handoff call in the target cell, the relative mobility in the SHO area

is periodically measured and estimated, as described previously. Once the relative

(a) (b)

Figure 3.4 Flow diagram of CCS construction and update

mobility is updated, the call will be checked if it stays in the NC-region of the target

cell such that psT(t,i)<psN(t,i). Then, all handoff calls in the NC-region should be

tested in the MSC to make sure that each of them holds at least two channels in its

Yes

Yes

Yes

No

No

No

Each accepted handoff call

Estimation of mobility

No. of channels in active set >1?

In the NC?

Pseudo handoff

call?

|CCS| = |CCS| + 1

End

No

No

Yes

No

Each handoff call in CCS

No. of channels in active set >1?

Call out or call completion?

Pseudo handoff call?

End

Estimation of mobility

|CCS| = |CCS| - 1

Yes

Yes

51

active set. cr_ps(t,i) is evaluated to check whether the call is stationary or moving

away from the target cell such that cr_ps(t,i)<0 or |cr_ps(t,i)|<ε. The call will be

added to the CCS of the target cell if all the constraint conditions in the CCS

definition are met.

On the other hand, Figure 3.4(b) illustrates how handoff calls are

removed from the CCS. Each handoff call in the CCS is also periodically checked.

The call will be taken out of the CCS of the target cell if one of following three

situations takes place: if it is out of coverage of the target cell, if it completes the

call, or if no longer satisfies the conditions for a pseudo handoff call.

3.3 CHANNEL CONVERSION WITH DYNAMIC GUARD CHANNEL

RESERVATION (CCDG)

3.3.1 Channel Allocation for SHO Requests

In order to improve the efficiency of channel utilization and reduce both

dropping and blocking probabilities, a new handoff scheme based on the relative

mobility of users is proposed. Calls in the NC-region of the target cell periodically

report their pilot strength to the BS and the MSC where the CCS is constructed and

updated as discussed in section 3.2.4.

Figure 3.5 shows the proposed handoff scheme for new handoff calls.

When a new handoff call arrives, the BS at the target cell first checks if there exists

any free channel. If so, the channel is allocated to the handoff call. If no free channel

is available and the new handoff call is a real handoff call, the noncontrolling

channel (or the channel with weaker pilot strength) of a call in the CCS is converted

to the new handoff request as long as the CCS is not empty. Otherwise, the new

handoff call is placed in a queue to wait for a free channel or a channel from a call in

the CCS. The call will be refused if the queue is full [94]. However, refusing a

handoff call request does not mean that it is dropped. A handoff call in the handoff

52

queue would be dropped only after the call moves out of the handoff area without

getting any channel from the target cell.

Figure 3.5 Flow diagram of the channel allocation in the CCDG scheme

3.3.2 Dynamic Allocation of Guard Channels

Since the dropping of a handoff call is considered more severe than the

blocking of a new call, a fixed number of channels are often reserved exclusively for

handoff calls. These exclusively reserved channels are referred to as guard channels.

It is not efficient for such guard channels to be reserved even when handoff traffic is

not heavy.

Thus in this scheme, the number of soft guard channels for handoff is

dynamically adjusted according to the variation of the CCS. The number of guard

channels is updated every time that the CCS is changed. The bigger the CCS, the

less the number of guard channels. Figure 3.6 shows a flow diagram of dynamic

channel reservation. At first, the number of guard channels is initialized as g0. Then,

No

No

No

Yes

Handoff call request HO

Free CH?

Estimation of mobility

|CCS|>0

Place HO into Queue Allocate CH to HO

The call dropped Convert a CH to HO

End

Pseudo call? Out of

handoff area?

No

Yes

Yes

Yes

53

the number of guard channels is dynamically adjusted. If the number of calls in the

CCS is less than or equal to 1, g remains the assigned value g0. If |CCS| > g0, which

means there are enough channels in the CCS that are available for new handoff calls,

g is set to 0. Otherwise, if 0 < |CCS| < g0, g is set to g0 − |CCS| + 1. This scheme

avoids unnecessary guard channel assignment when adequate channel resources are

available for handoff requirement.

Figure 3.6 Flow diagram of dynamic guard channel reservation

in the CCDG scheme

When channel conversion is performed, it is important not to influence

the quality of voice and increase the total interference. Since selection diversity is

used for uplink interference in a WCDMA system where one BS (called controlling

BS) that has a higher receiving power than another BS demodulates the received

signal, the transmitting power of the MS will remain almost the same. In the

proposed scheme, the noncontrolling channel of a call in the CCS is converted to an

incoming handoff call without remarkably degrading voice quality, handoff process

and interference requirement as the channel conversion goes on.

Yes

Allocate initial guard g0

|CCS| changes?

|CCS|>1?

|CCS|>g0?

g = g0 g = 0 g = g0 − |CCS|+ 1

End

No

Yes

No

Yes

No

54

3.4 RESULTS AND DISCUSSION

3.4.1 Simulation Parameters

The main parameters used in the simulation are given in Table 3.1.

Table 3.1 Simulation parameters for CCDG scheme

Td Number of channels in a cell

le Maximum handoff queue length

ah Ratio of the handoff area to the entire cell area

λn New call arrival rate in a cell

λh Handoff arrival rate in a cell

New call arrival rate in normal area

New call arrival rate in handoff area

Transferring rate of a call from the normal area to the handoff area

Transferring rate of a call from the handoff area to the normal area

Call departure rate from handoff area

λt Call departure rate from normal area

Moving rate to an adjacent cell

Moving rate from TC-region to NC-region of the target cell

Call departure rate from CCS

Call departure rate from handoff queue

Tc Channel holding time

µc Mean of channel holding time

c Probability that a handoff call is a pseudo handoff call

g Current number of guard channels

gO Predefined initial number of guard channels

Under the condition that all the neighbouring cells are statistically

identical and behave independently. The characteristics of the overall system can be

55

captured by focusing on a single cell [97]. In this analysis, atmost two different

sources in diversity reception are considered. Each cell will reserve g channels out

of a total of Td available channels exclusively for handoff calls (which are the guard

channels) [98]. Every handoff request is assumed to be perfectly detected in analysis

and the assignment of the channel is instantaneous if the channel is available. It is

also assumed that the allowable maximum handoff queue length is equal to le. In

addition, assuming that calls initiated within the cell arrive at a Poisson rate of ,

handoff request arrivals also form the Poisson process with rate and channel

holding time Tc follows an exponential distribution with mean μc-1. Assuming that

the location of a newly generated call is uniformly distributed over a cell and the

new call arrival rates in the normal and handoff regions are = (1 – ah) and

sλn = ah /2 respectively.

Consider that the new calls in the TC-region of the target cell as new

calls in the handoff region, from the viewpoint of the target cell, whereas the new

calls in the NC-region of the target cell are taken as handoff calls to the target cell.

The dwelling times of a call in the two distinct regions are assumed to be

exponentially distributed. The transferring rate of a call from the normal area to the

handoff area is and the transferring rate of a call from the handoff area to the

normal area is aλc . The rate that a call is terminated is denoted as λt. Numerical

results are reported in this section. Considering a system with parameters Td = 24,

ah = 0.3, g0 = 2 (channels), λt = 0.01 (calls/s), le = 4 and all call arrival rates are

assumed to be Poisson distributed. The simulation is done for the proposed scheme

and is compared with the conventional IS-95 SHO scheme [95].

3.4.2 New Call Blocking Probability

The new calls arising in a cell are blocked when all the available

channels are busy and it degrades the service offered by the network. In the

conventional IS-95 SHO scheme [96], the SHO calls occupy multiple channels and

are responsible for the blocking of new calls. In the proposed scheme, those extra

56

channels (weaker channels in the active set) are grouped into CCS and are available

for channel conversion for the new calls. These available channels in CCS reduce

the new call blocking probability which is evident from the Figure 3.7. The scheme

accommodates more handoff calls by channel conversion and gets more channel

resources for a new call due to dynamic guard channel adjustment.

3.4.3 Handoff Refused Probability

An incoming handoff request to a cell may be refused under either of the

following circumstances:

1) The handoff queue is already full or

2) It does not obtain a handoff channel until either call completion or

departure of the caller from the cell takes place.

Figure 3.8 depicts the handoff refused probability of the conventional

and the proposed scheme. The improvement on handoff refused probability by

CCDG handoff scheme is steady and significant and much greater than the

improvement made by channel borrowing [39] alone, since the CCDG scheme

Figure 3.7 New call blocking probability

57

distinguishes handoff calls in the CCS from ordinary handoff calls and serves more

handoff calls by channel conversion.

3.4.4 Average Number of Guard Channels

The average number of guard channels, which reflects the variation of

guard channel reservation for handoff call is equal to g0 for IS-95 handoff schemes.

Figure 3.9 shows the average number of guard channels for both the schemes. It is

constant for IS-95 handoff scheme because the number of guard channels in the

system is fixed throughout the channel allocation process. However, for the CCDG

handoff scheme, it decreases when a cell is in higher load which implies that the

channel conversion policy and dynamic channel reservation increase capacity for

new calls in the normal area. Since it is not efficient for guard channels to be

reserved even when handoff traffic is not heavy, the number of guard channels is

varied according to the variation of CCS. Number of guard channels is updated

every time that the CCS is changed. The bigger the CCS, the less is the number of

guard channels.

Figure 3.8 Handoff Refused Probability

58

3.4.5 Carried Traffic

The carried handoff traffic per cell is defined as the average number of

occupied channels in the handoff area of the target cell. The scheme makes more

handoff calls being served by the system. Total carried traffic increases as more

number of channels is available with increase in load. The increase in the traffic is

due to the fact that the proposed scheme is serving more channel resources (channels

in the CCS) when the load increases. Figure 3.10 depicts the carried handoff traffic

and Figure 3.11 illustrates the total carried traffic offered by both the schemes. It is

observed that there is significant increase in the traffic offered by the system in the

proposed scheme compared with that of the IS-95 scheme.

Figure 3.9 Average number of guard channels

59

Figure 3.10 Carried handoff traffic

Figure 3.11 Total Carried traffic

60

3.4.6 Channel Conversion Probability

Figure 3.12 shows the probability of channel conversion during the

handoff process for the CCDG scheme which increases with the call arrival rate.

However, as traffic is getting heavy, the increase of conversion probability gradually

turns slower till it finally becomes a constant. This is because more channels

occupied by handoff calls in the CCS may contribute to alleviating possible handoff

dropping as traffic is getting heavier and until the channel pool in the CCS is out of

supply.

3.4.7 Channel Efficiency

It is the ratio of the mean number of calls served in a cell to the total

carried traffic. The proposed scheme is more efficient than the conventional scheme

because of the increased channel resources (CCS and dynamic allocation of guard

channels). As discussed earlier, the proposed scheme offers more channel resources

Figure 3.12 Channel Conversion Probability

61

from the CCS hence when the load is getting heavier, those channels in CCS are

effectively converted for new requests. Further, the dynamic reservation of the

number of guard channels still enhances the performance of the system. Figure 3.13

shows the channel efficiency for both the schemes. It is found that the proposed

scheme is more efficient than the conventional IS-95 handoff scheme.

3.5 CONCLUSION

In this chapter, a new mobility-based SHO scheme is proposed and its

performance is analyzed with that of conventional IS-95/CDMA SHO scheme. It is

observed that there exist SHO calls that move away from the target cell to the

neighbouring cell or in stationary status which unnecessarily occupy multiple

channels. The proposed scheme discriminates such pseudo handoff calls from real

handoff calls by measuring and estimating their relative mobility, thus setting up a

CCS. When there are free channels, the handoff process operates in the same way as

Figure 3.13 Channel Efficiency

62

that in conventional IS-95/CDMA2000 cellular systems. However, when all

channels are occupied and a handoff request occurs, a weaker (or non controlling)

channel used by a call in the CCS is converted to a new handoff request.

The numerical results show that the proposed scheme outperforms the

conventional handoff scheme. Compared with the SHO schemes proposed in patents

and related publications, the advantages of the proposed scheme for SHO are that

1) The scheme can significantly reduce both new call blocking

probability and the handoff refused probability.

2) It does not noticeably degrade the QoS.