Benefits of Transmit and Receive Diversity in Enterprise Femtocell Deployments

Yi Jiang, Yan Zhou, Mohit Anand, Farhad Meshkati, Vinay Chande, Norman Ko and Mehmet Yavuz Qualcomm Inc.

5775 Morehouse, Dr., San Diego, CA 92121 USA E-mails: {yij, yanzhou, manand, fmeshkat, vchande, nko, myavuz}@qualcomm.com

Abstract—In this paper, we study the benefits of transmit and receive diversity for enterprise UMTS femtocell deployments. Indoor enterprise femtocell deployments face a single-path slow fading wireless environment that may lead to frequent hard handovers in the boundary region of neighboring femtocells and consequent degradation in the voice quality experienced by the users. In the absence of soft-handover (SHO) support, transmit diversity at the femtocell can combat single-path fading channel. We demonstrate through over-the-air tests that transmit diversity is very effective in reducing the number of hard handovers and therefore results in significant improvement in voice quality for enterprise users. On the uplink, we study system stability using an analytical approach. We derive analytical conditions for system stability with and without receive diversity at the femtocells. Using this analytical framework, benefits of receive diversity in maintaining system stability and preventing uplink power racing between neighboring femtocells are quantified. It is shown that, in the absence of SHO, receive diversity is very effective in maintaining system stability by preventing potential uplink power racing caused by inter-femto interference.

Keywords-femtocells; enterprise; transmit diversity; closed-loop transmit diveristy; handover; hard handover; receive diversity; system stability; power racing.

I. INTRODUCTION Femtocells are low-power cellular base stations typically

deployed indoors in residential, enterprise or hotspot settings.They offer multiple benefits to subscribers and operators. To subscribers, they promise excellent user experience at home (better coverage and higher data throughput) and access to specialized femtozone applications. Operators benefit from offloading traffic from macro cellular network and reduction of infrastructure and maintenance costs through unplanned deployments.

In this paper, we study the benefits of transmit and receive diversity for commercial indoor multi-femto deployments (i.e., enterprise femtocells). In enterprise femtocell deployments, due to the single-path slow fading indoor channel, a user equipment (UE) may experience frequent hard handovers (HHO) when in the boundary region of two neighboring femtocells1. Each HHO can result in several consecutive frame erasures, which can cause a noticeable voice artifact (NVA). Hence frequent HHOs can lead to noticeable degradation of user experience. Inter-femto soft handover (SHO) can mitigate this issue by providing

1 Since a large portion of UEs currently do not support Rx diversity, we assume absence of Rx diversity at the UE throughout this paper.

independent fading on each SHO leg and provide excellent voice quality. In absence of soft handover support (e.g., due to its implementation complexity), femtocell downlink transmit diversity (TD) provides an effective diversity method for overcoming the single-path fading channel. It is demonstrated in this paper via over-the-air (OTA) test results that TD reduces the number of HHOs and provides good voice quality for the enterprise users. As an additional benefit, the closed-loop transmit diversity (CLTD) provides coverage extension without increasing the femto transmit power. There are also other benefits of transmit diversity for femtocell deployment, e.g., inter-femto interference mitigation as studied in [1] (and the reference therein).

On the uplink (UL), receive diversity results in improved uplink throughput and reduction in the femto mobile’s transmit power. Reduction in femto UE Tx power reduces inter-femto and femto-to-macro interference. Inter-femto interference can potentially result in power racing between the users served by adjacent femtocells in an enterprise deployment. Specifically, without SHO, a Femto User Equipment (FUE) in the boundary region of two femtocells may increase its UL transmit power to react to deep fades caused by single-path fading indoor channel. The increase in the UE’s Tx power can significantly push up the UL interference to be above the noise floor at the neighboring femtocell. Due to the high interference, the FUE served by the neighboring femtocell will have to increase its transmit power, which will result in UL interference at the other femtocell. As a result, both FUEs enter power racing, hence increasing the noise rise at both femtocells and interference to the macrocell. As an effective solution, Rx diversity can be employed at the femtocells to mitigate power racing by reducing the UL fading and hence preventing sudden increases in FUE transmit power. To quantify the effectiveness of Rx diversity, an analytical framework is presented to identify the power racing condition. Based on the framework, the femto UL performance with Rx diversity is studied under various conditions. Numerical results show that Rx diversity can significantly reduce the power racing probability.

The remainder of the paper is organized as follows. In Section II, we study and verify the benefit of transmit diversity in reducing the number of hard handovers as well as in femto coverage extension. Section III identifies the UL inter-femto power racing condition and evaluates the effectiveness of femto Rx diversity in reducing power racing. Major conclusions are summarized in Section IV. The focus of this paper is 3G femtocells using UMTS/HSPA+ technology.

Indoor and Outdoor Femto Cells

978-1-61284-824-2/11/$26.00 ©2011 IEEE 456

II. VOICE QUALITY IMPROVEMENT VIA TRA

A. OTA Mobility Tests without Transmit DiverIn this section, we present over-the-air (OTillustrate that a UE at the boundary region of twexperience very frequent HHOs. Figure II-1 shows the layout of the OTwithin a building in Qualcomm campus. The retwo femtocells. In this test, the femto TDdisabled. A UE is placed at the boundary of theillustrated by the green circle. This location isthe UE has almost equal path loss from both femakes an AMR 12.2kbps voice call through one

Figure II-1 Layout of the HHO OTA

During the voice call, UE may experience two t

i. Conversational movement, e.g., thphone in talking position and with sand head movements.

ii. Walking movement where the testeforth between two femtocells in a speed of approximately 1km/hr.

The HHO happens when the serving cell andhave Ecp/Io (the ratio of the received piloreceived energy, also called Common PIlot CPICH Ec/Io) satisfying [2]

EI CIO EI hystwhere CIO stands for cell individual offset ofand hysteresis term is determined by configured

Figure II-2 shows the UE HHO behavior baselog. Here hysteresis-CIO=3dB. The top sufiltered Ecp/Io traces of the two femtocells waverage values. As seen in the plot, the Ecp/Isignificantly due to the single-path fading chain the middle illustrates the serving cell changWe see from the log that 9 HHO happens witest location at the boundary of two feconsecutive frame erasures can occur duriprocedure as illustrated in the bottom su

ANSMIT DIVERSITY

rsity TA) test results to wo femtocells may

TA test performed ed circles represent D functionality is e two femtocells as s chosen such that

femtocells. The UE e of the femtocells.

A test.

types of mobility:

he tester holds the some regular body

er walks back and 4 meter range at

d the neighbor cell ot energy to total CHannel Ec/Io or

teresis (2-1) f the neighbor cell, d triggering event.

ed on a typical test ubplot shows the which have similar Io values fluctuate annel. The subplot es, i.e., the HHOs. thin 19 sec in this

emtocells. Several ing the handover

ubplot. Therefore,

frequent HHOs may cause man(NVAs), which degrade the voicuser. For this enterprise building,floor falls into the handover regionone additional femto within 4dB of

Figure II-2 The OTA test logs

SHO can address many of the fading environment. The benefitsknown. It helps a voice call in providing diversity due to indepenproviding higher DPCH (dedicat(each of the SHO legs can transmit power), and (3) eliminating servicmake-before-break architecture. Itherefore, a desirable feature thaquality in a multi-femto deploymen

At the same time, SHO implementation complexity for entIn absence of SHO support, anotheto prevent frequent hard handovbetween neighboring femtocells. sections, transmit diversity at the fefor mitigating the impact of sinproviding good voice quality for en

B. Basics of Transmit Diversity

With 1Tx the CPICH received sigiven by RSCP T hwhere is the channel gain betweUE, and CPICH1 is the transmit CPa femto with TD, either the spaceloop TD (CLTD) [3], two orthotransmitted simultaneously [3]. TheRSCPTD h CPIC√

ny noticeable voice artifacts ce quality experienced by the about 20% of the enterprise

n where the UE can see at least f the strongest femto.

of UE HHOs without TD

issues arising in a single path s of soft handover are well multiple ways, namely - (1) dent fades on the SHO legs (2) ted physical channel) power the maximum allocated DPCH ce interruption artifacts due to Inter-femto soft handover is, at ensures smooth voice call nt.

support poses additional erprise femtocell deployments.

er source of diversity is needed vers at the boundary regions As shown in the following

emtocell is an effective method ngle-path fading channel and nterprise users.

ignal code power (RSCP) is

h CPICH n, (2-2)

en the femto antenna 1 and the PICH power by antenna 1. For e-time TD (STTD) or closed-ogonal CPICH sequences are e RSCP can be written as

CH h CPICH√ n, (2-3)

457

where CPICH and CPICH are orthogonal sequences and √2 in the denominator accounts for the transmission power reduction on each antenna so that the total transmit power remains the same as the 1Tx case. It is clear that the CPICH Ec/Io for the 1Tx system follows:

E TI | | (2-4) For the CPICH Ec/Io measurement of the TD system, the two CPICH channel qualities are measured separately and the final CPICH Ec/Io measurement is the combination of the two [4]. Hence for TD, the CPICH Ec/Io follows:

E TDI | | | | (2-5) Suppose the two transmit antennas have independent channel fading. It is clear from comparing (2-4) and (2-5) that with TD, CPICH Ec/Io benefits from diversity gain over the 1Tx case, although there is no beamforming gain since the mean values of RSCPTD and RSCP T , are the same.

C. OTA Test of Voice Quality with Transmit Diversity

Figure II-3 The OTA test logs of UE HHOs with TD

We now show the OTA test results with TD. The test setting is the same as the test shown in Figure II-2 except that here the TD functionality is enabled. The UE HHO behavior is illustrated in Figure II-3. Comparing the top subplots in Figure II-3 and Figure II-2, one can clearly see that the fluctuations in CPICH Ec/Io are significantly reduced with TD. Consequently, the HHO is much less frequent and so is the number of NVAs per minute. Table II-1 compares the number of HHOs occurred per minute with and without TD for different hysteresis settings. In general, the presence of TD reduces the HHO frequency significantly. Moreover, the benefit of HHO reduction due to TD is even more prominent when the hysteresis is set high. For hysteresis-CIO=3dB, the reduction in HHOs due to TD is about 30~50% while for hysteresis 6dB, the reduction in HHOs due to TD is about 80%. Hence combining TD with a reasonably high

hysteresis is an effective approach to reducing the HHOs and the associated NVAs, and achieving good voice quality

Table II-1 Number of HHOs per minute, 1 Tx antenna vs. TD

Hysteresis-CIO=3dB

Hysteresis-CIO=6dB

1Tx Conversational 30 11

1km/h walk 22 19

TD Conversational 14 2

1km/h walk 16 4

To conclude this section, we note that closed loop TD (CLTD) can provide range extension benefits in addition to diversity gain. CLTD, which relies on the feedback of beamforming vector via the UL Dedicated Physical Control CHannel (DPCCH) from the UE to the femto [3], provides both diversity gain and 3dB beamforming gain (in practice, however, the beamforming gain is less than 3dB due to the quantization of the beamforming vector and the time-varying channel). Hence given the same channel condition (measured by CPICH Ec/Io), the femto with CLTD is less likely to reach the Max Dedicated Physical CHannel (DPCH) allocation. This explains Figure II-4, which shows that given the same CPICH Ec/Io, CLTD achieves better frame error rate (FER) than 1Tx antenna.

Figure II-4 Coverage extension obtained by CLTD. CLTD can operate at lower frame error rate at given CPICH Ec/Io. The circles, triangles and squares represent OTA lab test results. The solid lines are trend lines. Max DPCH allocation is 10%.

III. UPLINK INTER-FEMTO POWER RACING MITIGATION VIA FEMTOCELL RECIEVE DIVERSITY

On the uplink, to achieve the same frame error rate, Rx diversity requires less average FUE transmit power by smoothing out the fading. The reduced transmit power results in less inter-femto and femto-to-macro uplink interference. Furthermore, in multi-femto enterprise deployments, the single-path slow fading nature of the indoor channel can cause inter-femto power racing on the UL. Without soft handover, Rx diversity can be very effective in mitigating power racing by preventing deep fades on the UL. In the following sections, an analytical framework is first presented to identify the power racing condition. Based on it, the benefit of femto Rx diversity in terms of power racing probability reduction under various conditions is quantified.

-20 -18 -16 -14 -12 -10 -8 -60

2

4

6

8

10

12

Average CPICH Ec/Io (dB)

FE

R %

Blue: 1Tx, Red: CLTD; DPCH Ec/Ior 10%

static testconverstationalfast movement

458

Figure III-1 Two-femto two-user model with FUE 1 and 2 served by

Femto 1 and 2, respectively.

A. Power Racing Condition with Single Rx Antenna In this section, the power racing condition for the case of

single Rx antenna at the femtocell is derived based on a simple 2-femto 2-user model in Figure III-1, where FUE 1 and 2 are served by femtocell 1 and 2, respectively. Each femtocell has one Rx antenna, and no inter-femto soft handover is assumed. The Signal to Interference plus noise Ratio (SIR) of FUE 1 at Femto 1 for a given channel realization can be written as

2212

1111

ngpgpsir

σ+=

where jp is the transmit power of the j-th FUE, jkg represents the channel coefficient from the j-th FUE to the k-th femto, and

2nσ is the noise power at each femto. The SIR of FUE 2 at Femto

2 can be derived similarly. Suppose the target SIR is jγ for the j-th FUE, the transmit power to achieve the target SIRs is given by the power solution to the following equations

⎩⎨⎧

==

22

11

γγ

sirsir

which can be rewritten in matrix form as

( ).

)13(

12212122112

21

2

11122

21122

22

121

22122

21111

2

1

gggggggg

gggg

pp

n

n

n

n

γγγσγσ

γγ

γσγσ

γγ

−⎥⎦

⎤⎢⎣

⎡⎥⎦

⎤⎢⎣

⎡=

−⎥⎦

⎤⎢⎣

⎡⎥⎦

⎤⎢⎣

⎡−

−=⎥

⎦

⎤⎢⎣

⎡−

The power racing will happen if no (finite) positive power vector exists that solves (3-1). This corresponds to both FUEs continuously increasing their transmit power as the power control can never achieve the target SIRs. It can be seen from (3-1) that the condition for no positive power vector is given by

212112

2211 γγ≤gggg

(3-2)

The above condition can be expressed in dB domain as

[ ] [ ] [ ] [ ][ ] [ ] 21

21122211

dBdB

dBdBdBdB ggggγγ +≤

−−+

which states that power racing will happen if the link gain difference is less than the sum of SIR targets. Therefore, the power racing will be more likely to happen for FUEs near the cell boundary. Furthermore, power racing can be mitigated by reducing target SIR to make the condition less likely to hold.

Finally, the power racing probability for one Rx is given by

⎟⎟⎠

⎞⎜⎜⎝

⎛≤= 21

2112

2211 γγggggprobwer racingprob of po

where the probability is over different channel realizations.

B. Power Racing Condition with Rx Diversity This section investigates the power racing condition with Rx

diversity at the femtocell. Following the 2-femto 2-user model in Figure III-1, the SIR of FUE 1 at Femto 1 in case of 2-Rx diversity with maximum ratio combining is given by

∑= +

=2

12

212

1111

i ni

i

gpgpsir

σ

where ijkg represents the channel coefficient from the j-th FUE

to the i-th Rx antenna at the k-th femto. For simplicity, the derivation assumes that the interference plus noise at the two Rx antennas are independent and Gaussian.

Same as the single Rx case, the power racing will happen if no positive power vector exists to achieve the target SIRs. However, finding an explicit power racing condition is challenging for Rx diversity case, since the SIR is not a linear function of transmit power. Therefore, we resort to a SIR lower bound to characterize the condition:

( )12

212

1112

1

2212

2

1111

1 rsiGp

Gp

gp

gpsir

n

in

i

i

i

′=+

=+

≥ΔΔ

=

=

∑

∑σσ

where ( ) .221jkjkjk ggG +=

Note that 1rsi ′ will become the actual SIR in case of one Rx. The transmit power to keep the SIR lower bound as the target SIR is the solution to the following equations

⎩⎨⎧

=′=′

.22

11

γγ

rsirsi

(3-3)

Similar to (3-2), the condition for no positive powers to solve (3-3) can be derived as

459

212112

2211 γγ≤GGGG

(3-4)

On the other hand, power racing will happen if and only if no positive power solution can be found to maintain the actual SIRs as the target SIRs. This implies that whenever power racing happens, there is also no positive power vector for keeping the SIR lower bounds as the target SIRs, which is a more stringent constraint. In other words, if actual SIRs cannot achieve target SIRs by varying powers, their lower bounds must be lower than target SIRs as well. Therefore, (3-4) serves as a necessary condition for power racing in case of 2 Rx antennas and will reduce to the sufficient and necessary condition (3-2) in case of 1 Rx antenna. By comparing (3-4) and (3-2), it can be seen that the power racing is less likely to happen with Rx diversity since the gain ratio is less likely to become small in the Rx diversity case due to the more stable effective channel gain jkG . Accordingly, the power racing probability can be upper bounded by:

.212112

2211⎟⎟⎠

⎞⎜⎜⎝

⎛≤≤ γγ

GGGGprobwer racingprob of po

The above upper bound will reduce to the actual power racing probability in case of one Rx antenna.

C. Numerical Results for Power Racing Probability In this section, the behavior of power racing probability is

studied based on the exact expression for 1 Rx and the upper bound for Rx diversity derived before. The validity of power racing condition is verified via detailed system-level simulations which model inner loop and outer loop power control. The results are not included here due to space limitation.

Figure III-2 Power racing probability behavior for different Rx

diversity configurations.

The benefit of Rx diversity in mitigating uplink power racing is evaluated in Figure III-2, where the performance is compared for different Rx diversity configurations including both 1 Rx and 2 Rx with spatial correlation of 0, 0.6, and 0.9. The 2-femto 2-user model in Figure III-1 is assumed with both FUEs having PL=65 and 75dB to the serving and non-serving femtocells, respectively. Rayleigh fading is assumed for each propagation link. It can be seen that Rx diversity significantly reduces the power racing probability. Furthermore, the similar performance for spatial correlation of 0.6 and 0 indicates that Rx diversity still achieves near optimal performance with antenna correlations typical to

femtocell environment. The mitigation of power racing is due to the fact that Rx diversity can reduce the chance of high FUE transmit power by preventing deep fades in the link to the serving femtocell. The impact of fading statistics on power racing probability is studied in Figure III-3, where each propagation link is assumed to have Rician fading with K dB power ratio of line-of-sight component to scattered component. Rx diversity with 0.6 antenna correlation is considered for this result, and both FUEs have PL=65 and 75dB to the serving and non-serving femtocells, respectively. It can be seen that the power racing probability drops dramatically with a higher K factor since deep fades become less likely.

Figure III-3 Power racing probability behavior for different Rician

K factors.

IV. CONCLUSIONS In this paper, we study the benefits of transmit and receive diversity at femtocells for enterprise femtocell deployments. It is shown that with femtocell transmit diversity (TD), a UE in boundary region of two neighboring femtocells experiences less signal fluctuations and hence less frequent hard handovers. Therefore, if inter-femto soft handover is not supported, transmit diversity can provide good voice quality in enterprise femtocell deployments. In addition to reducing number of HHOs, the closed-loop TD (CLTD) can provide range extension as well. On the uplink, Rx diversity at the femtocell provides improved uplink throughput and reduced femto mobile Tx power. The reduction in femto mobile’s Tx power results in lower uplink interference to macro and other femtocells. In addition to these benefits, we have shown that receive diversity at the femtocell is beneficial in mitigating UL power racing caused by inter-femto interference in enterprise deployments. We have obtained the power racing condition on the uplink and shown via numerical results that Rx diversity can significantly reduce the power racing probability. The effect of spatial correlation on the Rx diversity gains is minimal up to an antenna correlation factor of 0.6.

REFERENCES [1] M. Husso, J. Hämäläinen, R. Jäntti, J. Li, E. Mutafungwa, R. Wichman, Z.

Zheng, and A. M. Wyglinski, “Interference Mitigation by Practical Transmit Beamforming Methods in Closed Femtocells”, EURASIP Journal on Wireless Communications and Networking, Volume 2010 (2010)

[2] 3GPP TS25.331,v7.18.0, Radio Resouce Control (RRC); Protocol specification

[3] 3GPP TS25.211, v7.10.0, Physical channels and mapping of transport channels onto physical channels (FDD)

[4] 3GPP TS25.215, v7.4.0, Physical layer; Measurements (FDD)

-20 -15 -10 -5 0 5 1010

-5

10-4

10-3

10-2

10-1

100

Target SIR [dB]

Upp

er b

ound

of

pow

er r

acin

g pr

ob.

1 ant2 ant, 0 corr2 ant, 0.6 corr2 ant, 0.9 corr

-15 -10 -5 0 5 1010

-5

10-4

10-3

10-2

10-1

100

Target SIR [dB]

Upp

er b

ound

of

pow

er r

acin

g pr

ob.

K = -inf dBK = 1.5 dBK = 3 dBK = 6 dBK = 10 dB

460