Вы находитесь на странице: 1из 5

Evaluation of Inter-Frequency Quality Handover Criteria in E-UTRAN

Muhammad Kazmi1, Olof Sjbergh2, Walter Mller2 Wireless Access Networks, Ericsson Research1 Research and Development2 SE-164 80 Stockholm, Sweden {Muhammad.Kazmi, Olof.Sjobergh, Walter.Muller}.ericsson.com
Abstract In E-UTRAN multiple co-located carriers will typically be deployed requiring efficient inter-frequency (IF) handover procedures and algorithms for retaining service quality, better coverage and load balancing between the carriers. Five IF handover criteria using RSRP, RSRQ or combination thereof are investigated in synchronous and asynchronous EUTRAN deployment scenarios. The results depict that an IF handover solely based on RSRP significantly increases number of handovers. Conversely the handover criterion based only on RSRQ reduces handovers but it slightly increases the packet loss rate. The overall best performance is achieved with the combined handover criterion, which uses both RSRP and RSRQ and thus guarantees that the received pilot strength as well as the signal quality stays within the desired limit after the handover. Keywords; OFDMA, LTE, Radio Resource Management, Colocated carriers, RSRP, RSRQ, Inter-frequency HO.

Jonas Wiorek2 and Bengt Lindoff3 Ericsson Mobile Platforms, Ericsson Research3 SE-223 70 Lund, Sweden {Jonas.Wiorek, Bengt.Lindoff}.ericsson.com

II.

E-UTRA INTER-FREQUENCY HANDOVER SCENARIOS

To support various mobility scenarios in E-UTRAN two EUTRAN specific UE downlink measurement quantities are specified [3]: reference symbol received power (RSRP) and reference symbol received quality (RSRQ). The RSRP and RSRQ measurements are indeed analogous to WCDMA CPICH Ec/No and CPICH RSCP measurements respectively [4]. Thus RSRP is equivalent to the signal strength measurement and is defined as the linear average of the received power of the resource elements carrying cell-specific reference signals within the considered measurement frequency bandwidth. RSRQ is used to depict the cell quality and is defined as the ratio of RSRP to E-UTRA carrier received signal strength indicator (RSSI). The E-UTRA carrier RSSI is the linear average of the total received power in OFDM symbols containing the reference symbols. It should be noted that except the first OFDM symbol in a sub-frame, the remaining ones can also contain data resource elements [1] [5]. This means the E-UTRA carrier RSSI also incorporates the contributions from resource elements carrying the user data. This property of the E-UTRA carrier RSSI component in the denominator of RSRQ enables the depiction of the cell quality. An operator would typically deploy more than one EUTRA carrier frequency in the same coverage area. In order to ensure efficient use of multiple carrier frequencies deployed in the same coverage area, the 3GPP E-UTRAN standard provides necessary procedures, mechanisms and radio resource management requirements pertaining to the E-UTRA interfrequency handovers (IF) [2] [6]. In both E-UTRAN FDD and TDD the UE is required to perform RSRP and RSRQ measurements of at least 4 inter-frequency identified cells per E-UTRA carrier [6]. There is also a requirement on the UE to monitor up to at least 3 E-UTRA carriers [6]. This means in total an E-UTRA UE shall be capable of measuring at least 12 inter-frequency cells. An inter-frequency handover allows an operator to achieve one or more of the following objectives: retaining service quality, load balancing, retaining cell coverage etc. Our focus in this paper is to evaluate the first objective, that is, the performance of the quality based inter-frequency handover. In our analysis we make use of both signal strength (i.e. RSRP) and signal quality (i.e. RSRQ) to guarantee the service continuity while retaining the desired quality of service. This

I.

INTRODUCTION

Handover is one of the most fundamental radio resource management features in a mobile network. The E-UTRAN or the so-called 3GPP long term evolution (LTE) [1] supports mobility in several deployment scenarios: E-UTRA intrafrequency, E-UTRA inter-frequency and inter radio access technology (inter-RAT). The inter-RAT handover enables the mobility between E-UTRAN and other access technologies, which may comprise of WCDMA, GSM, High Rate Packet Data (HRPD) or cdma2000 1xRTT. In E-UTRAN the handover decision, which is taken by the serving cell, relies on the downlink measurements and/or the network configured events reported by the user equipment (UE). The evaluation of the handover related events specified in [2] are in turn based on one or more downlink measurements performed by the UE. To support inter-frequency and interRAT mobility scenarios the corresponding measurements and event evaluation are carried out by the UE during the network configured measurement gaps. Our goal is to devise and analyze suitable criteria for performing E-UTRA inter-frequency handover in E-UTRAN system. The E-UTRAN supports both time division duplex (TDD) and frequency division duplex (FDD) modes. Our analysis is though based on the E-UTRAN FDD but the conclusions related to the synchronous FDD scenario are also applicable to the E-UTRAN TDD mode.

978-1-4244-2517-4/09/$20.00 2009 IEEE

type of handover is needed when UE experiences poor downlink received quality due to the dominant co-channel interference from intra-frequency neighbour cells, which operate on the same carrier as that of the serving cell. This phenomena is commonly observed in high rise building environment, where despite strong received signal strength, the received signal quality can be poor due to the strong inter-cell interference originating from the neighbour cells. In such scenario an intra-frequency handover is less likely to improve the received quality since in other intra-frequency neighbour cells the UE is expected to experience similar level of interference. As shown in this paper, in such scenarios if two or more carriers are available an improvement in the performance can be achieved by employing a suitable inter-frequency handover scheme. It is also shown that under these circumstances the inter-frequency handover exhibits its benefits if both RSRP and RSRQ are used. This is because the use of these measurements ensures that the target inter-frequency cell has both better coverage and signal quality. A. Inter-frequency Handover Phases The inter-frequency (IF) handover procedure can be divided into the following four distinct phases: 1) Measurement gap triggering In this phase the measurement gaps are triggered according to an absolute threshold corresponding to one or more intrafrequency measurement quantities; for instance if the intrafrequency RSRQ falls bellow the network configured absolute threshold. It is therefore based on measurement(s) from the serving cell only. 2) Inter-frequency measurements and event evaluation In this phase the UE starts measuring inter-frequency cells on the target carrier frequency during the activated gaps. In addition the UE also performs intra-frequency measurement quantities in between the gaps in a usual manner. Based on these measurements the UE evaluates one or more network configured event. 3) Inter-frequency measurements and/or event reporting In this phase the UE reports to the serving cell both intrafrequency and IF measured quantities and/or the corresponding events evaluated during the previous phase. 4) Handover decision Based on the UE reported measurements and/or events the serving cell decides whether to perform the handover or nor by sending or not sending the handover command. The serving cell also chooses the most suitable cell in case multiple events are reported by the UE. B. Inter-frequency Handover Criteria As explained in the preceding section that inter-frequency handover evaluation should preferably be based on both RSRP and RSRQ. We analyze and compare various cases of interfrequency handovers based on different combinations of RSRP

and RSRQ as expressed in table 1. The set of symbols (;) and (1;2) represent absolute thresholds and handover margins respectively.
TABLE I. No 1 2 3 4 5 IF HANDOVER CRITERIA

Inter-frequency (IF) HO Evaluation Criteria IF measurement gap IF handover decision criteria triggering criteria RSRPserving_cell < (RSRPtarget_cell - RSRPserving_cell) > 1 RSRPserving_cell < RSRQserving_cell < RSRQserving_cell < (RSRPserving_cell < ) OR (RSRQserving_cell < ) (RSRQtarget_cell - RSRQserving_cell) > 2 (RSRQtarget_cell - RSRQserving_cell) > 2 (RSRPtarget_cell - RSRPserving_cell) > 1 (RSRPtarget_cell - RSRPserving_cell) > 1 AND (RSRQtarget_cell - RSRQserving_cell) > 2

III.

SIMULATION MODEL

The performance of the algorithms is evaluated using dynamic system simulation. The simulated area consists of 7 hexagonal cells per E-UTRA FDD carrier; 1 cell per site with site-to-site distance of 1.5 km. There are two E-UTRA FDD carrier frequencies (f1 and f2) belonging to E-UTRA band 1 (2.0 GHz range). A. Traffic Generation A mixed traffic scenario is considered in the simulation. The mixed traffic comprises of two types of users on both carriers in the system: Web surfing traffic Voice over IP (VoIP) traffic

Uneven traffic load is created on the two carriers; where f1 is deliberately more loaded than f2 by introducing more Web traffic on the former. The web traffic on the loaded carrier corresponds to around 50% resource utilization. The objective of introducing higher load on one of the carriers (i.e. f1) is to enable the activation of gap-assisted measurements and subsequent handover to a cell on f2 according to the evaluation criterion. Furthermore, there is one VoIP traffic user in the entire system whose performance is observed. The VoIP packet size comprises of 304 bits including all headers due to Real-time Transport Protocol (RTP), User Datagram Protocol (UDP) and Internet Protocol (IP). The generated mean VoIP bit rate is approximately 15.2 kbps. The voice packets which cannot be successfully delivered to the UE within 80 ms are dropped. B. Measurement Model The measurement bandwidth of RSRP and RSRQ comprises of 6 resource blocks (1.4 MHz). One transmit antenna at the eNode B and 2 receive antennas at the UE are used. Both receive antennas are assumed to be uncorrelated with equal gain. The physical layer (L1) measurement period for both intra-frequency RSRP/RSRQ and inter-frequency RSRP/RSRQ measurements is 200 ms. Though in E-UTRAN

standard [6] the inter-frequency L1 measurement period over which the minimum accuracy requirements of RSRP and RSRQ are to be met is 480 ms, but this is not likely to influence the conclusions of this paper. For each measurement quantity from each cell the UE obtains one snapshot of 4-5 ms in time domain once every 60 ms during the physical layer measurement period. This implies that 3 snapshots are averaged over 200 ms period to determine the L1 filtered RSRP and RSRQ. The eventual RSRP and RSRQ comprise of the linear average of the measured values from the two receiver diversity branches. Higher layer time-domain filtering, which is commonly known as layer 3 (L3) filtering [2], is used in dB scale. The L3 filter coefficient (k) is set to 4, which corresponds to the time constant of approximately 664 ms. Time to trigger, which is used for event evaluation is not configured. Handover margin (for both RSRP and RSRQ) is 1 dB. This means handover to the target cell target cell is considered provided the measurement quantity used for the event evaluation from the target cell is 1 dB stronger than that from the serving cell. C. Scheduling Strategy The resource blocks are assigned by the eNode B scheduler to the users in round-robin fashion both in time and frequency domains. This means in the frequency domain, the resource block allocation starts from the central resource block. The web traffic and VoIP users are scheduled with equal priority. D. Radio Environment The fast fading is used and the channel model consists of 3GPP-TU macro-cells [7]. The mean user speed is 90 km/hr. One transmit and two receive antennas are used at eNode B and UE respectively. The system bandwidth is 5 MHz. The Log-Normal shadow fading with standard deviation of 8 dB is used. The shadow fading correlation between cells is 0.5. IV.
SIMULATION RESULTS

The simulations are done in both synchronous and asynchronous networks. In the synchronous network scenario the frame timings of all the simulated cells are assumed to be perfectly aligned. In asynchronous network the frame timing of each cell is set independently. In accordance with the RSRQ definitions, in both synchronous and asynchronous scenarios, the E-UTRA RSSI part is measured in all the resource elements in OFDM symbols containing the reference signals. A. Synchronous Deployment 1) RSRP and RSRQ Variation Figure 1 and figure 2 show the variation of RSRP and RSRQ versus time respectively from the serving and one of the target cells in the simulated synchronous deployment scenario. Results are shown for two cases: low system load and high system load.
RSRP -85 -90 -95 -100 -105 dB -110 -115 -120 -125 -130 -135 Starting BS, high load Starting BS, low load Target BS, high load Target BS, low load 0 5 10 15 t [s] 20 25 30

Figure 1. RSRP variation in synchronous network


RSRQ 0

-2

In the simulations the voice user is not dropped due to the bad link quality instead a voice packet not delivered within 80 ms is discarded. Thus the performance of the voice users will be severely degraded if the correct type of handover is not timely executed i.e. the voice packet loss will increase. In order to ensure adequate voice reception quality the mean voice packet loss rate should not exceed 1%. Thus in order to judge the performance of the 5 different inter-frequency handover criteria mentioned in table 1, the mean voice packet loss rate is one of criteria used in our analysis. Furthermore, the most suitable handover criterion should on the average lead to lower number of handovers. Hence another criterion for the performance evaluation is the normalized mean number of handovers with respect to the reference case which is based on criteria # 1 (i.e. where only RSRP is used). In all scenarios the users perform the normal intrafrequency handover as well as the inter-frequency handover whenever necessary.

-4

dB

-6

-8 Starting BS, high load Starting BS, low load Target BS, high load Target BS, low load 0 5 10 15 t [s] 20 25 30

-10

-12

Figure 2. RSRQ variation in synchronous network

Figure 1 shows that the RSRP variation is insensitive to the system load. This is expected since RSRP is only measured on the resource elements containing only the reference signals, which are transmitted in the same manner regardless of the system load or user data. On the other hand figure 2 shows that the variation of RSRQ is significantly affected by the system load. This is due to the fact that the denominator in RSRQ (i.e. RSSI)

incorporates the contributions from the resource elements containing the data, which in turn reflects the cell quality experienced by the UE. Too large and fast variations in RSRQ are not desirable. Interestingly as observed in figure 2 the variations are not very severe. One reason is the use of higher layer time domain filtering leading to certain degree of stability in RSRQ measurement. Another possible reason is the use of round-robin scheduling in the simulations. 2) Mean packet loss rate Figure 3 shows the mean voice packet loss rate for all the five inter-frequency handover criteria in a synchronous network deployment scenario. The results illustrate that IF handover solely based on RSRP (criterion #1) does not even fulfill the target 1% voice packet loss rate. Similarly, criterion # 4 according to which the target cell is evaluated based on RSRP also leads to significantly higher voice packet loss rate. But criterion # 5 in which both RSRP and RSRQ are used, guarantees that the voice packet loss rate remains below the 1% target level.
Mean Voice Packet Loss Rate 2.5 Criteria #1 Criteria #2 Criteria #3 Criteria #4 Criteria #5

Relative Mean Number of IF HO 100 90 80 70 60 % 50 40 30 20 10 0 Criteria #1 Criteria #2 Criteria #3 Criteria #4 Criteria #5

Figure 4. Mean number of handovers in synchronous network

B. Asynchronous Deployment 1) RSRP and RSRQ Variation Figure 5 and figure 6 show the variation of RSRP and RSRQ versus time respectively from the serving and one of the target cells in the asynchronous scenario. Results for this scenario are also shown for the two cases: low and high system loads. Figure 5 shows that the RSRP variation is insensitive to the system load. This behaviour is very similar to the one observed in synchronous case. Thus as stated before, this is due to the fact that RSRP is only based on the reference signals, which are transmitted in the same manner regardless of the system load or user data.
RSRP -85

1.5 % 1 0.5 0

-90 -95 -100

Figure 3. Mean packet loss rate in synchronous network

Starting BS, high load Starting BS, low load Target BS, high load Target BS, low load

3) Mean number of handovers


dB

-105 -110 -115 -120 -125 -130 -135

Figure 4 shows the performance of mean number of handovers for all the five inter-frequency handover criteria in a synchronous network deployment scenario. The mean value is normalized with the reference case, which corresponds to criterion # 1. The results in figure 4 illustrate that although criteria # 4 ensures the lowest number of handovers compared to the reference case (criterion #1) but it also leads to the largest voice packet loss rate of 1% (as shown figure 3). The most desired handover criterion should ensure lower number of handovers as well as lower packet loss rate. Thus, from the perspective of the overall performance, we observe that criterion handover criterion # 5 (bases on both RSRP and RSRQ) exhibit the best performance.

10

15 t [s]

20

25

30

Figure 5. RSRP variation in asynchronous network

Figure 6 shows that the variation of RSRQ is significantly affected by the system load also in asynchronous scenario. This is again due to the E-UTRA carrier RSSI component in RSRQ, which reflects the cell quality experienced by the UE. The use of higher layer time domain filtering also ensures certain degree of stability in RSRQ measurement in this scenario.

RSRQ 0 -1 -2 -3 -4 dB -5 -6 -7 -8 -9 -10 Starting BS, high load Starting BS, low load Target BS, high load Target BS, low load 0 5 10 15 t [s] 20 25 30

(bases on both RSRP and RSRQ) also appears to be the best compromise in an asynchronous scenario.
Relative Mean Number of IF HO 100 90 80 70 60 % 50 40 30 20 10 0 Criteria #1 Criteria #2 Criteria #3 Criteria #4 Criteria #5

Figure 6. RSRQ variation in asynchronous network

2) Mean packet loss rate Figure 7 shows the mean voice packet loss rate for all the five inter-frequency handover criteria in an asynchronous network deployment scenario. The results illustrate that all five criteria tend to fulfill the target voice packet loss rate of 1%. However criteria # 2 and # 5 lead to relatively lower packet loss rate. This is because in both of these criteria the HO decision is based on at least the comparison of RSRQ between the serving and the target cells. This in turns ensure that the voice packet loss rate remain well below the desired limit.
Mean Voice Packet Loss Rate 2.5 Criteria #1 Criteria #2 Criteria #3 Criteria #4 Criteria #5

Figure 8. Mean number of handovers in asynchronous network

V.

CONCLUSIONS

System simulation results comparing several interfrequency handover evaluation criteria are shown in this paper. The two main performance measures are: mean voice packet loss rate and mean number of handovers. It has been shown that the IF handover scheme entirely based on RSRP leads to the worst overall performance compared to all other schemes, which involve RSRQ. Overall the handover criterion # 5 using both RSRP and RSRQ is observed to show the best performance. This is because the combined criterion (#5) ensures that after handover both signal quality and the pilot strength in the target cell are within the desired limit. The simulated scenario is comprised of 2 E-UTRA carrier frequencies, which are deployment in a regular hexagonal base station sites. In more challenging downlink limited scenarios, which are often encountered in high rise buildings in practical deployments, the RSRQ based handover schemes are likely to provide even more gain in terms of reduced number of IF handovers. Furthermore in such scenarios RSRP-only based IF handover would cause even higher packet loss rate since worse downlink quality is likely to be experienced due to the delay in the handover process.

1.5 % 1 0.5 0 0

REFERENCES
[1] [2] [3] [4] [5] [6] [7] E. Dahlman, S. Parkvall, J. Skold, Johan, P. Beming, 3G Evolution: HSPA and LTE for Mobile Broadband. Academic Press Inc, 2008. 3GPP TS 36.331, Evolved Universal Terrestrial Radio Access (EUTRA); Radio Resource Control (RRC); Protocol specification. 3GPP TS 36.214, Evolved Universal Terrestrial Radio Access (EUTRA); Physical layer - Measurements. 3GPP TS 25.215, Physical layer Measurements (FDD). 3GPP TS 36.211, Evolved Universal Terrestrial Radio Access (EUTRA); Physical Channels and Modulations. 3GPP TS 36.133, Evolved Universal Terrestrial Radio Access (EUTRA); Requirements for support of radio resource management. 3GPP TR 25.943, Deployment aspects, version 4.1.0.

Figure 7. Mean packet loss rate in asynchronous network

3) Mean number of handovers Figure 8 illustrates the normalized mean number of handovers for the five inter-frequency handover criteria in an asynchronous network deployment scenario. The figure shows that although criteria # 4 lead to the lowest number of handovers but the voice packet loss rate was highest. This is because the unnecessary avoidance or delay in the handover increases the packet loss rate. As expressed earlier, the most desired handover criterion should ensure lower number of handovers as well as lower packet loss rate. Thus, in terms of the overall performance, the handover criterion # 5

Вам также может понравиться