1. Introduction

Open access has been regarded [1] as a feasible solution to the problem of cross-tier interference in two-tier networks. Nevertheless, open access femtocell deployments are hardly practical due to the elevated number of required handovers. Indeed, when outdoor users are allowed to connect to any available cell (i.e., macrocell or femtocell) it is likely that due to the nomadic nature of these users, their connections would be continuously transferred between adjacent femtocells, or between femtocells and the umbrella macrocell. Furthermore, it is also well known [2] that HOs are not always successful and connections might be dropped as a consequence of HO failure. Additionally, the excessive signaling that emanates from an open access femtocell tier increases the complexity of the access network and introduces the need for large and powerful femtocell gateways.

1.1. Terminology

In order to clarify the concepts used throughout this article, the terminology to be applied is presented in the following. First, the main femtocell access policies are described.

In addition, in a closed and hybrid access, mobile users are classified as follows.

Moreover, the types of interference in two-tier networks are classified as follows.

Finally, note that variables in lower case represent magnitudes in natural units, while upper case indicates logarithmic scale, that is, dB.

1.2. Related Work

In order to cope with the previously mentioned problems, automatic pilot power control has been proposed in [4] as a possible solution for Code Division Multiple Access (CDMA) femtocells. However, this approach could lead to insufficient indoor coverage. In similar way, dynamic antenna patterns have also been suggested in [5], at the expense of a slightly more complex hardware in the FAPs. Furthermore in [6], a decentralized OFDMA resource allocation scheme was presented that optimizes the Area Spectral Efficiency (ASE) by applying spectrum fragmentation. However, this approach limits the maximum achievable instantaneous throughput regardless of the interference.

1.3. Contribution

It is because of the previously described drawbacks of open access that the closed access femtocell approach still seems appealing to most mobile network operators. Hence, novel solutions to the interference problem caused by CSG femtocells are still needed. In this article, a HO and interference mitigation technique based on the use of IHO for OFDMA femtocells operating in CSG mode is proposed. IHOs are widely used in Global System for Mobile communication (GSM) networks, where users are changed from channels with low-signal quality to channels that are in better conditions. The objective of this work is to apply the same concept to OFDMA subchannels, similarly as is done in GSM, in order to mitigate cross-tier interference.

In Section 2, interference is described in the context of two-tier networks. In Section 3, the notation used and several key concepts are presented. In Section 4, the IHO approach proposed in this article is depicted. In Section 5, the dynamic system-level simulation used in order to verify the performance of the IHO approach is summarized. In Section 6, a performance comparison in terms of network outages and throughput between closed, open access and IHO is given. Finally, in Section 7, the conclusions are drawn.

2. Interference in Two-Tier Networks

In two-tier networks, the severity of interference depends on two main factors: the strategy used for allocating the spectral resources to the tiers and the method used for accessing the femtocells [7]. A discussion on this topic follows.

2.1. Impact of the Spectrum Assignment on Network Performance

2.1.1. Assignment of Spectral Bands

Operators having more than one licensed spectral band have the following options for spectrum allocation [8].

It is to be noticed that deploying a two-tier network using more than one carrier introduces a notable problem of battery drain in the User Equipment (UE), especially in the dedicated and partially shared approaches. In this case, femtocell subscribers connected to the macrocell tier perform a continuous search for the femtocell carrier, which is highly energy-consuming from the radio-interface point of view. Once the femtocell subscribers find the femtocell carrier, they synchronize to it and check, in the case of CSG access, their connectivity rights. Furthermore, if the femtocell subscriber is not allowed to connect to the CSG femtocell, its UE must resynchronize to the macrocell carrier, which supposes a new search, and therefore a further battery consumption.

Since not all operators have more than one spectral band to divide between tiers, and because the shared spectrum approach is more challenging from the technical point of view (it results in a better spectral efficiency, at the expense of having to mitigate cross-tier interference), the rest of this article focuses on the single carrier case.

2.1.2. Assignment of OFDMA Subcarriers

In OFDMA, the spectrum is divided into orthogonal subcarriers that are then bundled into groups called subchannels (Wireless Interoperability for Microwave Access (WiMAX)) or resource blocks (LTE). Therefore, operators owning one spectral band and deploying a two-tier OFDMA network have different choices to allocate subchannels to the macrocell and femtocell tiers [9].

It must also be mentioned that in [10], a hybrid approach has been proposed in which femtocells far from a macrocell use a cochannel assignment, whereas the close femtocells apply an orthogonal approach.

As explained in the previous section and in a similar way to the shared spectrum approach, a cochannel assignment of subchannels results in a larger spectral efficiency as long as cross- and cotier interference are efficiently mitigated, for example, by using self-organization techniques. To illustrate this, Table 1 shows the results of a system-level simulation of a two-tier OFDMA network (scenario shown in Figure 3) using the self-organization approach presented in [11]. The experiment verifies that the performance of a cochannel assignment in terms of network throughput is better than that of an orthogonal assignment, mainly due to the better frequency reuse.

Table 1: Performance comparison.

2.2. Impact of the Access Method on Network Performance

2.2.1. Interference in Closed Access

With closed access, nonsubscribers can receive severe jamming from nearby femtocells. Even if the femtocell pilot power is larger than that of the nearest macrocell, nonsubscribers are not allowed to use the femtocell and are thus interfered in the downlink. Moreover, nonsubscribers transmitting with high power can cause interference in the uplink of nearby femtocells. The most challenging case of cross-tier interference, in this case, occurs when a nonsubscriber enters a house hosting a CSG femtocell. Then, in the downlink, the interference from the FAP is much stronger than the macrocell carrier, thus jamming the visitor. Similarly, the visitor can jam the uplink of the FAP.

2.2.2. Interference in Open Access

With this access method, nonsubscribers can also connect to femtocells. Therefore, the problem of a nonsubscriber passing by or entering a house where a femtocell is deployed is nonexistent. Hence, open access reduces the impact of cross-tier interference, which can be verified by the experiments presented in Table 2. However, open access has two major drawbacks.

Table 2: Performance comparison.

Furthermore, in large deployments (high femtocell densities) of open access femtocells, even if a nonsubscriber is connected to a femtocell, the aggregate of all cotier interference coming from neighboring femtocells can disrupt its service (Table 2). It is to be noticed that cotier interference is also a problem in closed access.

As a conclusion, it is to be noticed that both access methods have drawbacks: CSG increases cross-tier interference, whereas open access increases the number of HOs [12]. In this article, the use of Intracell Handovers (IHOs) is proposed in order to cope with both issues.

3. Preliminaries

In the following, the notation used in the rest of the article is presented. Moreover, the concepts of neighboring cell list, measurement report, channel quality indicator, and handover are introduced.

3.1. Neighboring Cell List and Measurement Report

In order to select the best serving cell when the UE is idle, or to aid the HO procedure when the UE is active, the UE measures continuously the RSSs of the pilot channels of the neighboring cells. In order to simplify and speed up the task of the UE when monitoring the air interface, the serving cell periodically broadcasts to its UEs the list of cells and pilot channels that they must measure. This list is known as the Neighboring Cell List (NCL). After receiving the NCL, the UE performs (every period of duration ) the appropriate measurements, and reports back the results to its serving cell, which then decides whether to start a new HO procedure or to take no action.

In two-tier networks, the NCL of a macrocell not only contains neighboring macrocells, but also open femtocells. Therefore, nonsubscribers must report back the RSSs of the pilot channels not only from all neighboring macrocells, but also from open femtocells.

Nevertheless, the case of CSG femtocells is different from open access. In order to minimize the impact of femtocell deployments on the existing macrocell tier, macrocells do not provide information about CSG femtocells in their neighboring cell list to the UEs. However, UEs perform an autonomous search to detect CSG femtocells [13].

3.2. Channel Quality Indicator

By using the Channel Quality Indicator (CQI), a UE can also report periodically, for example, at most every in LTE networks, to its serving cell its signal quality in terms of Signal to Interference plus Noise Ratio (SINR), as well as the signal qualities of a given subset of resource blocks (usually the ones with better conditions). This CQI is used by the Medium Access Control (MAC) layer for channel-dependent scheduling and rate control, but it can also be used to trigger a HO when the UE reports a low SINR.

Furthermore, let us mention that in several wireless standards, such as WiMAX and LTE, the user equipment has the capability of estimating the instantaneous SINR in all subcarriers [14, 15].

3.3. Network Definition

Let us define a two-tier OFDMA network as a set of:

where the NCL of cell contains different cells (macrocells and femtocells) and it is denoted by .

3.4. Handover

In cellular networks, a HO is triggered when the RSS () of the pilot signal from a serving cell at a mobile is lower than that of a neighboring cell (). These signal strength measurements are signal levels averaged over time using measurement reports. This averaging is necessary in order to cope with fading. Moreover, when a mobile user is located in between cells, it could happen that its transmission is ping-ponged from cell to cell. In order to avoid this effect, a hysteresis margin for the HO decision is also used. Furthermore, an umbrella cell system is normally deployed to minimize the large number of HOs incurred by high-speed users.

Then, if the HO condition is met, the serving cell establishes communication with the target cell and a HO is performed. However, in two-tier networks, performing a HO is not always the best choice or a possible option at all, because

In order to overcome these drawbacks, and the limitations imposed by closed and open access (CSG increases cross-tier interference, whereas open access increases the number of HO), the following IHO approach is proposed.

4. Intracell Handover in Femtocell Networks

The IHO is a special type of HO in which the source and target cell is the same one. The purpose of IHO is to transfer a user from a channel, which may be interfered or faded, to a clearer or less-faded channel. IHO is used, for example in GSM networks, where it is triggered when a UE reports a large RSS, but a low Received Signal Quality (RSQ) due to interference and/or fading.

The main idea of the proposed approach is that when a nonsubscriber that is connected to a macrocell suffers from cross-tier interference due to a nearby femtocell, the macrocell itself performs an IHO if possible, or casts an IHO in all interfering femtocells otherwise.

Without loss of generality respect to the IHO approach, the rest of this article focuses only on the DownLink (DL).

4.1. Triggering an IHO

When the SINR of a nonsubscriber connected to macrocell in subchannel is smaller than a given threshold , the serving macrocell requests a measurement report from nonsubscriber . In this measurement report, nonsubscriber indicates the RSS of its serving macrocell in subchannel (), as well as the RSSs of its neighboring cells in such subchannel ().

Note that the neighboring cells of are identified by the NCL provided by macrocell to nonsubscriber , and that they can be both macrocells and femtocells. Nevertheless, for the sake of simplicity it is assumed that the macrocell tier has been deployed using network planning and optimization tools, for example, base station location [16], automatic frequency planning [17], and thus macrocell intercell interference can be disregarded.

After receiving this measurement report, macrocell compares its with the other reported . Thereafter, macrocell triggers an IHO only if condition (1) is verified by a neighboring cell , and this neighboring cell is a femtocell .

where denotes an interference protection margin. If is too low, the nonsubscriber may suffer from interference before the IHO is launched, and its service could be dropped. Contrarily, if is too large, the IHO may be triggered to solve a nonexistent problem, thus increasing the signaling overhead in the network. Therefore, must be carefully selected in order to launch the IHO before the nonsubscriber falls into outage, while minimizing the signaling overhead.

Thus, if femtocell verifies (1), it is considered as a cross-tier interferer of nonsubscriber in subchannel , and therefore an IHO is triggered. Note that the set of all interfering femtocells of nonsubscriber verifying (1) is denoted by .

Thereafter launching the IHO, it can be performed either in the serving macrocell or in all interfering femtocells . In order to minimize the signaling overhead, it is preferred to perform this IHO in the serving macrocell than in all interfering femtocells . However, this is not always possible due to traffic load or interference conditions in the macrocell .

In the following, the conditions that are taken into account to decide whether an IHO is performed in the serving macrocell or in all interfering femtocells are presented, as well as the taken actions.

4.2. Performing the IHO in the Macrocell

An IHO is launched in the serving macrocell only if

In the case that there are more than one available subchannels in macrocell , nonsubscriber is switched to the subchannel that suffers the least interference. This subchannel is selected by macrocell according to the CQIs reported by nonsubscriber . In order to do this, macrocell can instruct nonsubscriber to perform measurements on unused subchannels, and report back their signal quality over the pilot reference symbols using the CQI.

It is to be mentioned that if no subchannel fulfills these requirements, the IHO is not performed in the serving macrocell , but in all interfering femtocells .

4.3. Performing the IHO in the Femtocells

If the IHO cannot be performed in the macrocell, it is attempted in all interfering femtocells verifying condition (1). In this case, the macrocell establishes communication with these femtocells (Section 4.4) and initiates an IHO in them. The signaling overhead caused by these communications is analyzed in Section 6.5. Moreover, and according to the availability of subchannels in the interfering femtocell , the IHO procedure differs. When there are available subchannels in femtocell , a subchannel switching process is carried out, whereas when there are no available subchannels in femtocell , a power control procedure is executed. Figure 1 summarizes this sequence of actions. Both procedures are introduced next.

Figure 1: Intracell Handover approach.

4.3.1. Switching Approach for IHO in the Femtocells

Switching is only possible when there is at least one available subchannel in the interfering femtocell . Then, the femtocell subscriber , which is currently connected to subchannel , is transferred to subchannel . In this way, subchannel is liberated, thus avoiding cross-tier interference to nonsubscriber . Note that if there are more than one available subchannels in femtocell , subscriber is switched to the subchannel with best conditions according to its CQIs. In order to do this, femtocell can instruct subscriber to perform measurements on unused subchannels, and report back their signal quality over the pilot reference symbols using the CQI.

4.3.2. Power Control for IHO in the Femtocells

If there are no free subchannels in the interfering femtocell , then one option would be to disconnect from subchannel for a period of time in order to avoid cross-tier interference towards nonsubscriber . However, this would decrease the throughput of femtocell , which is undesired. Note that this approach, called subchannel forbidding, is not part of the proposed IHO algorithm, but will be used for comparison in the following.

Then, instead of disconnecting from subchannel , the power transmitted by in subchannel is reduced for a period of time . In this way, cross-tier interference towards the nonsubscriber is mitigated.

The primary objective of this power control algorithm (illustrated in Figure 2) is to manage cross-tier interference, while the secondary objective is to maximize the femtocell throughput. In this way, the reduction of the throughput of femtocell is minimized compared to that of the forbidding approach presented above.

Figure 2: Power control algorithm for IHO.

Figure 3: Received power per OFDMA subcarrier (OFDMA-512 system) in a residential area covered by 1 macrocell and 64 femtocells. Each femtocell premises contains several indoor users, while 8 pedestrian users are located outdoors, being their routes also indicated.

The target of this distributed power control is to set the power , with which all interfering femtocells transmit in subchannel , to a value that ensures a certain signal quality to nonsubscriber .

In order to guarantee such , the maximum interference that nonsubscriber can tolerate is

where denotes the background noise.

Then, macrocell asks all interfering femtocells to decrease their transmit power in subchannel from to so that is respected for nonsubscriber . Furthermore and in order to avoid unfair power decrease requests among femtocells, the power decrease is weighted by macrocell using the RSS reported by nonsubscriber from each interfering femtocell according to

where is the maximum interference that femtocell is allowed to cause to nonsubscriber and is the path loss between them. Then, (3) simplifies to

where

being the power reduction in decibels requested for subchannel , which is computed by macrocell and passed on (Section 4.4) to all interfering femtocells .

Finally, it can occur that if femtocell is already transmitting with too little power in subchannel or if is too high, the power decrease request might have the same effect as switching off subchannel . In this case, subchannel is forbidden in femtocell for a period of duration , avoiding cross-tier interference towards nonsubscriber . However, it must be mentioned that subscriber is only disconnected from subchannel if it can afford it, that is, has allocated more subchannels or carries a service where delay is not crucial, for example, best effort, nonreal-time service.

Let us also indicate that the decrease in the femtocell throughput when forbidding a subchannel is statistically small, since this type of IHO is performed only from time to time when the femtocell is fully loaded and a nonsubscriber passes by the femtocell proximities.

Moreover, it is worth mentioning that a femtocell can generally afford to liberate a subchannel during a short period of time to avoid the outage of a mobile outdoor nonsubscriber. For instance, a WiMAX femtocell with 5MHz bandwidth can achieve in the downlink. Since a femtocell can only be accessed simultaneously by at most 4 users and subchannel forbidding is only triggered when the femtocell is fully loaded (otherwise, there would be available subchannels to switch to), then in order to fully use all available subchannels, it is likely that the femtocell users are not only using VoIP ( kbps) or video ( kbps), but also other intensive best effort services, for example, peer to peer. Then, they can afford to free a subchannel for a short period of time in favor of avoiding a nonsubscriber outage.

4.4. Macrocell and Femtocell Communication

In order to initiate an IHO in a femtocell, the macrocell needs to communicate with it. This communication could be established over the following.

Given the state of the art of mobility management in LTE femtocells, and because no interface has been standardized yet for HeNodeBs, the exchange of messages between macrocells and femtocells is more likely to be implemented over the backhaul connection. This can be done in a similar way as proposed approaches for macro to femto and femto to macro HOs, using the femtocell Gateway (GW) [19].

Since this communication is out of the scope of this paper, it is not analyzed in detail in the following sections. However, the signaling overhead originated in the system due to the IHO approach is analyzed in Section 6.5.

5. Experimental Setup

To evaluate the performance of the proposed IHO approach, dynamic system-level simulations have been used.

The scenario under scrutiny is a residential area with a size of in the town of Luton (U.K.), containing 438 premises of which around 400 are dwelling houses. 64 of these were selected to potentially host an indoor femtocell. Assuming that 3 operators with equal customer share provide services in this area, these 64 femtocells represents an approximate 50% femtocell penetration. Besides these femtocells, the scenario also contains one outdoor macrocell. Note that experiments were carried out with different femtocell penetrations: 50% (= 64 femtocells), 25% (= 32 femtocells) and 12.5% (=16 femtocells). The setup with 64 femtocells is illustrated in Figure 3, while the parameters of the simulation are shown in Table 3.

Table 3: Simulation parameters.

In this experimental evaluation, the different types of customers follow a well-defined behavior.

5.1. Propagation Model

For the path-loss predictions two different propagation models were used

The Root Mean Square Errors (RMSEs) of the macrocell and femtocell models are and , respectively.

5.2. System Level Simulation

An event-driven dynamic system-level simulation is used to model the operation of this two-tier OFDMA network. In this case, the life of the network through time is modeled as a series of events. For instance, an event happens when a user changes its position, when an IHO is launched, and so forth.

Because traffic modeling is out of the scope of this article, it is assumed that a user is allocated in each OFDMA frame to only 1 subchannel having OFDM symbols. Under this assumption, the SINR and throughput per user and other statistics are computed at regular time intervals.

This system-level simulation supports Adaptive Modulation and Coding (AMC). The different Radio Access Bearers (RABs), together with their SINR thresholds and efficiencies are shown in Table 4. Note that a UE cannot transmit when its SINR is lower than the SINR threshold of the lowest RAB defined in the network, that is, .

Table 4: RABs (Modulation and coding schemes).

Following the behavior of real-time services, a user is considered to be in outage when it cannot transmit for a given period of time. In this case, this period of time is fixed to as recommended by [23] for VoIP applications.

Further information about this system-level simulator such as channel modeling, interference modeling, throughput calculation, and so forth. can be found in [11].

5.3. Closed and Open Access Implementation

In the closed access simulations, nonsubscribers are always connected to the macrocell. These are likely to suffer from outage when they pass close to a femtocell that is making use of the same subchannel. It is assumed that after this type of outage, the nonsubscriber reestablishes its connection or call as soon as its SINR is larger than the SINR threshold of the lowest RAB defined in the network.

In open access simulations, outdoor users are always connected to the best server regardless of whether it is the macrocell or a femtocell. This case, outdoor users send a measurement report to their serving cell based on its NCL on a regular basis (). This indicates the RSSs of pilot channels of neighboring cells. Then, a hard HO is performed if the RSS of the strongest neighboring cell is larger than the one of the current server. It is to be noted that the HO procedure is carried out by the network and that there is a 2% probability that it fails, resulting in a dropped call. Then, it is assumed that after this type of outage, a nonsubscriber reestablishes connection immediately.

6. Results

First of all, let us mention that the IHO threshold has been set to , which is a value slightly larger than that of the lowest RAB defined in the system, that is, . In this way, the IHO is launched before the user cannot achieve any RAB, and it falls into outage, that is, the user cannot transmit for more than .

6.1. Effect of Parameter

Table 5 summarizes the network performance in terms of the outage of nonsubscribers, when using IHOs with different values of the interference protection margin . The simulations have been carried out with different number of femtocells in order to analyze the impact of the femtocell penetration on . It has been observed that if is low, for example, , a large number of outages occur. The reason is that the IHOs are launched either too late, when the nonsubscriber is already in outage, or they are not even launched. Moreover, when the femtocell penetration grows, the interference protection margin needed to fight outages is larger. For example, is sufficient in the case of a 25% penetration, whereas is needed in the case of a 50% penetration. This is due to the fact that with larger femtocell penetrations, the aggregate of cotier interference from neighboring femtocells towards a nonsubscriber grows. Therefore, the IHO must be launched when lower increases in interference are detected. Finally, it is worth mentioning that if is too large, an IHO will be launched to try to solve a problem (cross-tier interference) that does not even exist, increasing thus signaling, which is undesired.

Table 5: Outages of nonsubscribers.

In view of the results of Table 5, is assumed in the rest of the article because it is seen to guarantee outage avoidance in all cases.

6.2. Effect of Parameter

Figure 4 shows the sample scenario used to illustrate the performance in terms of SINR of a nonsubscriber, when it moves across the scenario. In this case, the performance of the IHO approach with different values of is compared to that of closed access.

Figure 4: Performance comparison of CSG and IHO for a mobile nonsubscriber ().

First of all, let us note that in closed access, when a nonsubscriber moves close to femtocells A or B (Figure 4(a)), it is jammed due to cross-tier interference (Figure 4(b)). This is assuming that femtocells A and B are using all available subchannels. Thus, the nonsubscriber falls into outage, since its SINR is smaller than that of the minimum RAB defined for a period longer than .

However, when using the IHO approach and the SINR of the nonsubscriber decreases, an IHO is launched in both femtocells A and B. Then, these femtocells stop using the subchannel used by the nonsubscriber. In this way, cross-tier interference towards the nonsubscriber is mitigated, and the outage is thus avoided.

Moreover, it can be seen in Figures 4(b) and 4(c) that if is not finely tuned, femtocells A and B begin to use the forbidden subchannel before the nonsubscriber moves out of their coverage area. In this case, new IHO must be launched in the femtocells in order to avoid the nonsubscriber outage, increasing thus the signaling overhead in the network.

However, if is finely tuned (see Figure 4(d)), femtocells A and B will not use the forbidden subchannel until the nonsubscriber is out of their coverage domain. Since a pedestrian walks at an average speed of and because the femtocell radius is estimated to be [22], has been proven to be appropriate to force only one IHO per femtocell in a residential scenario with mobile pedestrians.

6.3. Effect of Power Control

Figure 5 illustrates the performance in terms of SINR of the nonsubscriber moving along the route defined in Figure 4(a). The IHO approach with and without power control is compared here to closed access. In this case, the macrocell and femtocells are fully loaded. Therefore, an IHO based on subchannel switching is not possible.

Figure 5: SINR level of a mobile nonsubscriber.

When the IHO approach is applied without power control, that is forbidding interfering femtocells from using the subchannel employed by the nonsubscriber, the SINR of such nonsubscriber grows notably due to the avoidance of cross-tier interference (Figure 5(a)). However, this is at the expense of reducing the femtocell throughput, since a subchannel is forbidden (Figure 5(b)).

Contrarily, when the IHO approach is used with power control, subchannels are not forbidden, but the power applied to it is reduced. This is done in a controlled manner in order to protect the nonsubscriber. As a result, the SINR of the nonsubscriber is not as large as when forbidding the subchannel, but the outage is avoided and the femtocell throughput is enhanced with respect to the forbidding case. Figure 5(b) shows a case in which power control does not recover the full throughput capacity compared to CSG, but provides a gain with respect to the forbidding approach. In this case, this throughput gain is about  kbps, which is enough to hold real-time services such as VoIP ( kbps).

6.4. Closed, Open and IHO Comparison

In this section, the performance of the IHO approach compared to that of the closed and open access is analyzed. This has been done under different femtocell penetrations (12,5%, 25%, and 50%), and traffic loads in both the macrocell and femtocells (50% (4 users) and 100% (8 users)). There are 8 subchannels available for transmission.

The different setups are as follows.

Setup
Macrocell and femtocells are both half loaded (4 users). Under these conditions, IHOs based on subchannel switching are mostly launched in the macrocell (see Table 6 for results).

Table 6: Performance comparison for Setup 1 (800 s simulation).

Setup
Macrocell is fully loaded (8 users), whereas the femtocells are half loaded (4 users). IHOs based on subchannel switching are mostly launched in the femtocells (see Table 7 for results).

Table 7: Performance comparison for Setup 2 (800 s simulation).

Setup
Macrocell and femtocells fully loaded (8 users) and IHO implemented without power control. This case, subchannels are forbidden in the femtocells (see Table 8).

Table 8: Performance comparison for Setup 3 (800 s simulation).

Setup
Macrocell and femtocells fully loaded (8 users) and IHO implemented with power control under different for the nonsubscribers (see Table 9).

Table 9: Performance Comparison for Setup 4 (800 s simulation).

The analysis follows next with regard to different performance metrics.

6.4.1. Number of HO and IHO Attempts

First of all, let us note that in the CSG case, HOs are not allowed. Then, this value is neglected in the result tables.

It is to be mentioned that in all setups with open access, the number of HOs increases with the femtocell penetration. The more femtocells a nonsubscriber finds along its route, the more hand-ins and hand-outs must be carried out. Furthermore, the number of HOs also increases with the number of nonsubscribers.

Similarly, the number of IHOs also increases with the femtocell penetration and the nonsubscriber density. However, the number of IHOs launched is significantly less than the triggered HOs for open access in the same period of time. The reason behind this is that a unique IHO is needed to mitigate the cross-tier interference coming from one femtocell, while in open access two HOs are required (one hand in and one hand out). In addition, the HO is done based on the pilot signal, which is always transmitted, while the IHO only happens if the nonsubscriber and the interferer are utilizing the same subchannel (it does not always occur). Moreover, in this case and due to the multi-path effects, the coverage provided by the femtocells is not continuous. As a consequence, a nonsubscriber moving at low motion can hand over several times from the macrocell to the same open femtocell and vice versa. In order to mitigate this effect, a HO margin, which ensures that the HO is performed only if the neighboring cell is stronger by a given threshold than the server, could be considered. Nevertheless, in this way, the outages due to cross-tier interference would increase. In this case, a perfect HO is considered (as many HOs as needed are done), since the target is to compare open access and the IHO approach based on the number of outages, but not on the signaling overhead.

6.4.2. Outages Due to Interference

In all setups, closed access deployments are severely affected by cross-tier interference, resulting this in a large number of outages. As soon as a nonsubscriber walks near a femtocell using the same subchannel, the nonsubscriber falls into outage. This confirms the need of novel approaches for interference avoidance in two-tier networks.

In the case of open access, the number of outages due to interference are notably reduced compared to that of closed access. This is because nonsubscribers are allowed to connect to the strongest cell, turning the strong interferer into their best server. However, when the femtocell penetration grows, some cases of outage due to interference also appear. This is because even if the nonsubscriber is connected to the strongest femtocell, the aggregate of cotier interference from neighboring femtocells can disrupt its service, thus creating outage. It is to be noted that the signal strength of a femtocell outdoors is in the same order of magnitude as the signal strength of a neighboring femtocell located few meters away and therefore, if the femtocell penetration is large, cotier interference is a problem.

In the case of IHO, however, the number of outages due to interference is always zero. The number of outages, in this case, does not depend on the femtocell penetration, since the IHO approach is able to “switch off” all existing interferers independently of their number and position.

6.4.3. Outages Due to HO Failure

In open access, the number of HO failures increases with the femtocell penetration and the nonsubscriber density. Let us remember that according to [2], there is a 2% probability that a HO attempt results in a dropped call (outage). The more HO attempts, the more HO failures. Nevertheless, the number of outages due to HO failure in open access is much smaller than due to interference in CSG.

6.4.4. NonSubscriber Throughput

In all setups, open access deployments achieve a larger throughput for nonsubscribers than closed access. This is because in open access, nonsubscribers always connect to the strongest cell (macrocell or femtocell), which mitigates cross-tier interference and provides larger RABs. Moreover, the total nonsubscriber throughput in the IHO approach is also always larger than that of closed access due to cross-tier interference mitigation.

When comparing open access and the IHO approach, two issues must be highlighted.

The reason behind this is that in the IHO approach, in the first case, the macrocell just changes the interfered subchannel of the nonsubscriber to another with less interference. However, in the second case, all interfering femtocells stop using such subchannel, thus increasing notably the signal quality and throughput of the nonsubscriber.

It is also to be noticed that in all setups and for all techniques, the larger the femtocell penetration, the lower the total nonsubscriber throughput. This is again due to interference reasons: the more femtocells, the larger the interference, and therefore, the lower the SINR of the nonsubscribers, which is thus translated into a reduction of the nonsubscribers throughput.

6.4.5. Subscribers Throughput

Open access deployments achieve a larger throughput for subscribers than closed access and the IHO approach. This is due to the fact that when a nonsubscriber connects to a femtocell, a subchannel is released in the macrocell. This way, cross-tier interference towards subscribers is reduced, and the throughput of femtocells close to the macrocell is enhanced.

Let us now compare closed access with IHO.

The reason is that when the IHO is based on subchannel forbidding or power control, the interfering subchannel is either banned from femtocell use or its allocated power is reduced for a period of time. This causes a throughput loss for the subscribers. On the contrary, in closed access, subscribers are always connected to the same subchannel (no power variation) and thus their throughput is not reduced.

It is also to be noticed that in all setups and for all techniques, the larger the femtocell penetration, the larger the total subscriber throughput. This is because the more femtocells, the larger number of connected users in the femtocell tier. However, the average throughput per subscriber does not depend on the femtocell penetration, as it does in the nonsubscribers tier. This is because the effects of cotier interference indoors are smaller than outdoors, namely due to the shield provided by the house walls.

6.4.6. Power Control Behavior

Table 9 presents the performance of IHO with subchannel forbidding compared to that of IHO with power control. Table 9 also shows results for IHO with power control using different values.

This experimental evaluation shows, in agreement with the results presented in Figure 5(b), that when power control is used, the reduction of the femtocell throughput is less compared to that of subchannel forbidding. However, it occurs at the expense of decreasing the nonsubscribers throughput, since the interfering subchannels are not turned off.

It is also to be mentioned that the lower the value, the lower the reduction of the femtocell throughput. However, it occurs at the expense of reducing the nonsubscriber throughput, since they are only allowed to achieve lower signal qualities.

6.5. Signaling Overhead Due to IHO

The IHO approach implies the following signalin.

The size of a measurement report is

where denotes the number of interfering femtocells (), indicates the number of bits required to encode the identity of a neighboring cell, and represents the number of bits required to encode the RSS of a neighboring cell.

The size of an IHO message is

where denotes the number of bits required to encode the identifier of the subchannel, while , and are the number of bits required to encode and , respectively.

Let us assume that 3 bits are used to encode (there are 8 subchannels), while 8 bits ( levels) are used for , , and . Therefore, the number of required bits per measurement report and IHO message are 512 and 19, respectively. Note that an average of 32 interfering femtocells are considered in this calculation, which is the maximum NCL size in UMTS networks.

The measurement report is triggered when the SINR of a nonsubscriber connected to macrocell in subchannel is smaller than a given threshold . In the worst case scenario (Setup 4 (Table 9)), when having a femtocell penetration of 50% and using and , an average of 23.38 IHOs have been triggered per femtocell (simulation time = ). Therefore

These values are negligible compared to the capacity of the downlink and uplink of current OFDMA standards such as LTE or WiMAX. Therefore, it can be concluded that only a small fraction of the whole available bandwidth is needed for signaling overhead.

Finally, it is to be noted that the signaling required for an IHO is lower than for a HO. When performing a HO, all packets stored in the source cell, which belong to the user that is to be handed over, has to be transferred from the source cell to the target cell (implying a large overhead). However, in the IHO approach, in the worst case scenario, the macrocell has to indicate to the interfering femtocells only the reduction of power that has to be applied to a given subchannel and for how long.

7. Conclusions

The results of the system-level simulations show the following evidence about this type of residential scenarios.

Conclusion 1. The main problems of open access are the risk of outage due to HO failure and the high signaling introduced in the network. On the other hand, CSG femtocells introduce serious jamming problems (dropped call) to macrocell users in the downlink.

Conclusion 2. Open access has always an overall better throughput performance than the one of closed access.

Conclusion 3. When the femtocell penetration is large, the aggregate of the interference of nearby femtocells might be enough to cause outage to outdoor users, who will not be able to take advantage of the open access.

Then, it has been shown that by applying an intracell handover approach to OFDMA two-tier networks, the following occurs.

Conclusion 4. Downlink cross-tier interference to nonsubscribers decreases compared to CSG deployments. This is made evident from the number of outages due to interference shown in Tables 6, 7, 8, and 9 (results).

Conclusion 5. Handover attempts and thus network signaling decrease with respect to the open access case. Furthermore, the risk of outage due to handover failure is removed as it can be seen in the tables of results.

Conclusion 6. In order to avoid macrocell deterioration due to femtocell deployment, FAPs must be flexible when limiting the throughput of their own subscribers. The interference reduction of nonsubscribers should thus have more priority than the maximization of subscribers throughput. Therefore, a trade-off between these two objectives must be always achieved. Moreover, due to the nature of the services used by subscribers, the impact of subchannel forbidding is, in average, not too crucial for femtocell connectivity.

Conclusion 7. In order to reduce signaling, IHOs are attempted in the macrocell before they are attempted in nearby femtocells. However, it has been observed that femtocell-based IHOs result in higher throughput gains than macrocell ones. This is because a femtocell IHO removes fully the cross-tier interference while a macrocell IHO selects a less interfered subchannel which is not necessarily free of cross-tier interference.

Conclusion 8. Power control helps to handle cross-tier interference, while limiting the impact to subscribers when subchannel switching is not possible.

To summarize, the IHO approach presented in this article has better performance than CSG in terms of outage and throughput of nonsubscribers in all tested femtocell penetration conditions. The throughput of nonsubscribers is slightly below than that of open access in most cases, except for large femtocell penetrations, in which the IHO approach outperforms open access (Table 8). However, the network signaling is lower in case of IHO, and thus the risk of HO failure is practically nonexistent, avoiding outage.

Acknowledgments

This work is supported by the EU FP6 project on 3G/4G Wireless Network Design “RANPLAN-HEC” with Grant no. MEST-CT-2005-020958 and EU FP6 “GAWIND” with Grant no. MTKD-CT-2006-042783.