mobility parameter planning for 3gpp lte: basic concepts and

Mobility Parameter Planning for 3GPP LTE: Basic Concepts andIntra-Layer MobilityJari Salo

This white paper discusses the design of intra-layer mobility parameters for 3GPP LTE radio networks. Firstly,basic UE measurements defined in 3GPP Release 9 are reviewed. A measurement example illustrates their basicdifferences. Both idle mode and connected mode measurements and cell change rules from 3GPP are summarizedin a vendor-independent way. Secondly, design criteria for setting handover margin, time to trigger and measure-ment filtering coefficient are discussed in detail. A simple parameter design example is given. The scope of thispaper is limited to a review basic of concepts and design of intra-layer (intra-frequency) mobility parameters.

1. INTRODUCTION

This technical white paper introduces idle andconnected mode mobility parameter design for3GPP LTE. The target audience are radio plan-ning and optimization engineers with some ex-perience in LTE. Principles of OFDMA and SC-FDMA, as described in 3GPP LTE specifications,will not be repeated in this paper. Instead thereader is referred to well-known literature ref-erences [1–3]. Both FDD and TDD variants ofLTE radio interface are addressed and the dif-ference between the two is highlighted, wherenecessary.

In Section 2 Reference Signal Received Power(RSRP) and Reference Signal Received Quality(RSRQ) are defined and their properties dis-cussed. The relation of RSRQ to signal-to-noise-plus-interference ratio (SINR) is shown, andcomplemented by a measurement example il-lustrating the properties and interdependenciesof the different RF measurement quantities. Tokeep the paper self-contained, Section 3 sum-marizes the idle mode mobility rules for 3GPPRel9, while Section 4 does the same for con-nected mode handovers; the treatment is keptas vendor-independent as possible. In the keysection of this paper we outline criteria on howto design handover parameters for practical net-work deployment. The selection of handover

margin, time to trigger and L3 filtering coef-ficient is a trade-off between network interfer-ence, probability of unnecessary ("ping-pong")handovers and too late handover triggering. Fi-nally, a design example summarizes the keypoints of the paper.

The primary goal of the paper is to introducesome ideas for planning intra-layer mobility pa-rameters. Post-launch mobility parameter tun-ing based on network measurements are verycase-specific topics and beyond the scope here.

Version: 16 June 2013

1

2 J. Salo

Abbreviations

BCCH Broadcast ChannelCPICH Common Pilot ChannelFDD Frequency Division MultiplexingLTE Long Term EvolutionNAS Non-Access StratumOAM Operation and MaintenancePCH Paging ChannelPCI Physical Cell IdentityPDSCH Physical Downlink Shared ChannelPRB Physical Resource BlockPSS Primary Synchronization SignalRAT Radio Access TechnologyRE Resource ElementRRC Radio Resource ControlRS Reference SignalRSCP Received Signal Code PowerRSRQ Reference Signal Received QualityRSRP Reference Signal Received SignalRSSI Received Signal Strength IndicatorSIB System Information BlockSINR Signal-to-Interference-Noise-RatioSNR Signal-to-Noise RatioSSS Secondary Synchronization SignalTDD Time Division MultiplexingTTI Transmission Time IntervalUE User EquipmentUSIM UMTS Subscriber Identity Module

2. DEFINITION OF RSRP AND RSRQ

In this section, RSRP and RSRQ are defined,including their relation to SINR. A measure-ment example is shown to illustrate the mainpoints.

2.1. Reference Signal Received Power (RSRP)RSRP is defined as

RSRP =1K

K

∑k=1

Prs,k , (1)

where Prs,k is the estimated received power (inWatts) of the kth Reference Signal Resource El-ement transmitted from the first BTS antennaport. In Figure 1 these REs are denoted with

R0. To improve the accuracy of the RSRP esti-mate, the UE may optionally also measure theRS transmitted from the second antenna port(R1), if present. In case of four BTS transmit an-tennas, Reference Signals of the third and fourthBTS antenna ports are not used in the RSRPmeasurement. Since all LTE UEs have at leasttwo receive antennas, it is also mandated by [6]that the RSRP must be equal or higher than thestronger of the two receive antennas’ individualmeasured RSRP.

The maximum number of PRBs over whichRSRP should be measured1 is sent to the UEover RRC signalling, and is denoted in thispaper with Nprb . For example, in case of10MHz measurement bandwidth (Nprb = 50),the OFDM symbol carrying R0 (Figure 1) con-tains a total of 100 REs for R0, hence in this caseK = 100 in Eq. (1) for this OFDM symbol2 as-suming UE implementation does not use R1 inRSRP estimation.

RSRP measures only the RS power and ex-cludes all noise and interference power. ThusRSRP is a purely coverage-based handover trig-ger which is ideally independent of the networkload, similar to CPICH RSCP in 3G. It shouldbe noted that, since it is defined as the averagepower of RS resource elements, RSRP does notdepend on the number of BTS transmit anten-nas.

The reporting range of RSRP is divided in 1dBbins over −140 . . . − 44 dBm, the lowest RSRPvalue having reported value of ’0’.

The 3GPP specification does not specify howthe RSRP measurement should be implementedin a UE. However, the RSRP measurement accu-racy requirements are stated in some detail in[5], see Table 1 for a summary. Under normaloperating conditions, absolute measurement ac-

1Measurement bandwidth can be less than the actual down-link system bandwidth. This is useful for keeping measure-ment parameter configuration simple in multi-bandwidthdeployment scenarios.23GPP specification does not state any requirements on howmany OFDM symbols UE should average. Only measure-ment accuracy requirement is stated.

Mobility Parameter Planning for 3GPP LTE 3

0R

0R

0R

0R

0R

0R

0R

0R

1R

1R

1R

1R

1R

1R

1R

1R

an O

FD

M sy

mb

ol

con

tainin

g R

0

subframe

Freq

Time

Figure 1. Reference Signals as seen by UE,shown for transmit antenna port 1 (R0) andtransmit antenna port 2 (R1) for one 1ms sub-frame and 12 subcarriers.

curacy is allowed to have up to 6dB error forintra-frequency RSRP measurement. MeasuredRSRP difference between serving and a neigh-bour cell on the same carrier frequency shouldhave at most 3dB error. For inter-frequencyRSRP measurement, the absolute and relativeerror under normal conditions should both beless than 6dB. It can be seen that inter-frequencyRSRP measurement is considerably less accu-rate for "power budget" type handover trigger-ing than intra-frequency RSRP measurement.

The above cited measurement accuracy re-quirements hold for SNR = −6dB and withoutany higher layer measurement filtering. HigherSNR and application of L3 filtering results inimproved measurement accuracy. Therefore, intypical network conditions, measurement errorcan be expected to be smaller than shown in Ta-ble 1.

Table 1UE RF measurement L1 accuracy requirements.

Parameter AbsoluteAccuracy⋆

RelativeAccuracy⋆

RSRP intra-freq ±6dB ±3dBRSRP inter-freq ±6dB ±6dBRSRQ intra-freq ±3.5dB n/aRSRQ inter-freq ±3.5dB ±4dB

⋆ When SNR ≥ −6dB, RSRP≥ −127 . . . − 124dBmdepending on frequency band [5].

2.2. Reference Signal Received Quality(RSRQ)

Reference Signal Received Quality is the ra-tio of RSRP and total received signal and noisepower normalized to 1PRB bandwidth. As a for-mula we have

RSRQ = NprbRSRPRSSI

(2)

= Nprb

1K ∑K

k=1 Prs,k

∑Nren=1 Pn

, (3)

where Nre = 12Nprb is the total number of REs(including RS REs) in the measurement band-width Nprb, and Pn is the total received poweron the nth RE. The RSSI term in the denomima-tor of Eq. (3) is the sum power received over theOFDM symbol that contains R0, including owncell power, thermal noise and interference fromother cells. For example, for 10MHz (50PRBs)measurement bandwidth Nre = 600 and, assum-ing that only R0 is measured, K = 100.

As with RSRP, implementation of the RSRQmeasurement is not defined in detail by 3GPP.Measurement accuracy requirements are givenin Table 1. It can be seen that no rela-tive accuracy requirement has been defined forintra-frequency RSRQ. This is because in intra-frequency case the ratio of RSRQs depends onlyon the ratio of RSRPs of serving and neighbourcells. This can be seen by writing down theRSRQ ratio between two cells on the same car-

4 J. Salo

rier frequency, that is,

RSRQserv

RSRQneigh=

NprbRSRPserv

RSSI

NprbRSRPneigh

RSSI

=RSRPserv

RSRPneigh, (4)

since RSSI does not depend on the measured cellas it is simply the total received signal power3.Therefore, the ratio of RSRQs only depends onthe ratio of the RSRPs. This essentially makesrelative RSRQ redundant as a trigger for intra-frequency handover. In contrast, ratio of RSRQsmay be a useful trigger for inter-frequency han-dover where it can be used to direct UEs to aless loaded layer.

The reporting range of RSRQ is divided in0.5dB bins over −19.5 . . . − 3 dB, the lowestRSRQ value having reporting value ’0’. For un-loaded cell having one transmit antenna port,we have RSRQ= −3dB, assuming no other-cellinterference and negligible thermal noise. Simi-larly, a cell with two transmit antenna ports hasRSRQ= −6dB. On the other hand a fully loaded1Tx cell has RSRQ≈ −11dB, again neglect-ing any neighbour cell interference and ther-mal noise. Therefore, from mobility parame-ter planning viewpoint, setting any RSRQ-basedhandover trigger to a value higher than −11dBcould result in "ping-ponging" if UE happens tomeasure RSRQ at a time instant when there isdata download ongoing in the serving cell. Be-cause of this dependency on the instantaneoustraffic load, RSRQ measurements should be av-eraged over a longer period than RSRP. This isvery similar to the fluctuation of CPICH Ec/N0seen in HSDPA networks.

2.3. Relation of RSRQ and SINRIt is a common practice to use Signal-to-

Interference Ratio (SINR) as an indicator fornetwork quality. It should be however notedthat 3GPP specifications do not define SINR and3Since RSRQ is in general measured at different time in-stants for the serving and the neighbour cell, instantaneousRSSI could actually be different depending on fading vari-ations and subcarrier activity during the measured sub-frames. Time-averaged RSSI could however be assumed ap-proximately equal for serving and neighbour cell.

therefore UE does not report SINR to the net-work. SINR is still internally measured by mostUEs and recorded by drive test tools. Unfor-tunately UE chipset and RF scanner manufac-turers have implemented SINR measurement invarious different ways which in the authors’field experience are not always easily compa-rable. While at first it may seem that defin-ing SINR should be unambiguous, in case ofLTE downlink this is not the case. This is be-cause different REs within a radio frame carrydifferent physical signals and channels each ofwhich, in turn, see different interference powerdepending on inter-cell radio frame synchro-nization. For example, in a frame-synchronizednetwork SINR estimation based on synchroniza-tion signals (PSS/SSS) results in different SINRthan SINR estimation based on Reference Sig-nals, since in the latter case the frequency shiftof the RS depends on the PCI plan. This is illus-trated in Figure 3, and will be discussed in moredetail in the sequel.

In what follows we show one way of convert-ing RSRQ to SINR. Towards this end, we firstdefine average subcarrier SINR as

SINR =RSRPI + N

, (5)

where I is the average interference power andN is the thermal noise power. All quantitiesare measured over the same bandwidth and nor-malized to one subcarrier bandwidth. In OFDMown cell interference is often assumed to be neg-ligible and consequently I is due to other cellinterference only. RSSI reads

RSSI = Stot + Itot + Ntot , (6)

where the subscript ’tot’ indicates that thepower is measured over the 12Nprb subcarriersof the measurement bandwidth. The total serv-ing cell received power depends on the numberof transmitted subcarriers in the OFDM symbolcarrying R0, and on the number of transmit an-


tennas4. We can model this impact using theper-antenna subcarrier activity factor ρ and set

Stot = ρ12NprbRSRP . (7)

The value of ρ = 1 indicates full load such thatall subcarriers of one transmit antenna are trans-mitted for the OFDM symbol carrying R0. Ifonly RS is transmitted (i.e., unloaded cell) theresulting subcarrier activity factors would beρ = 1

6 and ρ = 13 for one and two transmit

antennas, respectively; Figure 1 illustrates this.When calculating ρ for two transmit antennas,one should take into account that REs overlap-ping with adjacent antenna RS transmission aremuted, and therefore, for example, in a fullyloaded 2Tx cell the scaling factor is ρ = 5

3 , in-stead of two5.

Since subcarrier interference plus noise poweris

I + N =Itot + Ntot

12Nprb, (8)

by combining the above equations we have thefollowing relation of average subcarrier SINRand RSRQ

SINR =1

112RSRQ − ρ

(9)

These formulas are illustrated in Figure 2. Anuncomfortable property of the RSRQ to SINRmapping is that it depends on the instantaneousserving cell subcarrier activity factor ρ, which istypically not known in live network measure-ments. Another problem is the sensitivity ofthe mapping on RSRQ values; a small changein RSRQ can result in a very large change inSINR which makes such mapping difficult to

4For the first OFDM symbol of the subframe this includesRS, PDCCH, PHICH, PCFICH subcarriers. For other OFDMsymbols this may include PDSCH, PSS/SSS and PBCH sub-carriers, depending on the subframe. Note that ρ may varyeven between different OFDM symbols of the same sub-frame.5It is assumed that all subcarriers have the same power, i.e.,there is no power boosting for any channel.

use in fading radio conditions. Indeed, plottinga RSRQ versus SINR scatter graph from a drivetest measurement one rarely obtains such a nice-looking curve as shown in Figure 2.

2.4. Measurement ExampleThe RF quantities discussed in this section are

illustrated by the scanner measurement shownin Figure 3. The scanner location is roughly be-tween two sectors of the same BTS; note thatthe serving and neighbour cell RSRPs that arewithin two decibels of each other. In the be-ginning of the measurement, both cells are fullyloaded (all subcarriers are transmitted). After 85seconds, the neighbour cell PDSCH utilizationdrops to 0%. At 165 seconds, also the servingcell PDSCH load6 drops to 0%. The followingobservations summarize the main points in thissection:

• RSRP is independent of own cell load andneighbour cell interference. This illus-trates how, from handover triggering pointof view, it simply measures coverage simi-larly to CPICH RSCP in 3G.

• RSRQ reacts to load changes in both serv-ing cell and neighbour cell.

• At the end of the measurement, RSRQ ≈−9dB. This agrees with the theoretical ( 1

8 )value when the UE is at cell border of twounloaded cells.

• The RSRQ difference between the two cellsis about the same as the RSRP differ-ence. Hence using relative RSRQ as intra-frequency handover trigger is redundant:using relative RSRP alone is sufficient withthe added benefit that RSRP measurementis also more robust under bursty networktraffic. The relative RSRP can be realizedusing the A3 measurement event, to be de-scribed later in the paper.

6BCCH and PCH also use PDSCH, so strictly speakingPDSCH utilization will always be slightly above 0%.

6 J. Salo

−20 −18 −16 −14 −12 −10 −8 −6 −4−10

−5

0

5

10

15

20

25RSRQ versus SINR for different serving cell loads

RSRQ (dB)

SIN

R (

dB)

1Tx unloaded1Tx fully loaded2Tx unloaded2Tx fully loaded

Figure 2. Theoretical comparison of RSRQ versus SINR.


• Since the radio frames are transmitted atthe same frame timing in both cells, theSecondary Synchronization Signals (SSS)overlap fully in time. Therefore SINR mea-sured from SSS is continuously close to0dB, independently of cell loads. In TD-LTE all cells in the network are frame-synchronized, which makes SSS SINR asimple and useful indicator for measuringphysical RF dominance.

• Serving cell load does not impact servingcell RS SINR since PDSCH of the servingcell does not overlap with serving cell RS.On the other hand, neighbour cell RS SINRis reduced by serving cell load since serv-ing cell PDSCH overlaps with neighbourcell RS7.

• In the beginning of the measurement,where both cells are fully loaded RSRQ ≈−15dB and SINR ≈ 0dB, which corre-sponds to Figure 2 for fully loaded 2Txcell.

• At the end of the measurement, whereboth cells are idle, serving cell RS SINR ≈15dB. Ideally, the RS SINR should be muchhigher since the reference signals of thetwo cells do not overlap and only ther-mal noise limits RS SINR. The exact reasonfor the RS SINR degradation is unclear,but might be caused by neighbour cell PD-CCH colliding with own cell RS. The the-oretical SINR in Eq. (9) is defined for thewhole measurement bandwidth, while inthe scanner measurement shown in Figure2 SINR is apparently measured based onRS resource elements only (the SINR esti-mation method used was not documentedby the scanner vendor).

7In the measurement, the serving and neighbour cell areframe-synchronized and the RSs of these two cells do notcollide because of the PCI plan.

3. INTRA-FREQUENCY IDLE MODE MO-BILITY

3.1. Introduction3GPP Release 8 specifies idle mode cell rese-

lection rules, which allows defining an absolutepriority for a cell; usually all macro cells of afrequency are assigned the same priority. Basedon these so-called priority-based cell reselectionrules, the UE always selects the network layer(cell) that has the highest absolute priority, pro-vided that the RSRP and RSRQ from the cell areabove a configured threshold value8. This mech-anism is different from the conventional pre-Rel8 cell reselection where the UE reselects toa cell with the best signal level or signal quality,instead of priority. In case the layer priorities areequal the signal level and quality still determinethe serving cell even in 3GPP Rel9. The lengthytopic of inter-layer mobility is beyond the scopeof this white paper. In this section, we focus onthe intra-layer case where all cells are assumedto have equal priority.

The term ’layer’ in this paper means a setof cells having the same carrier frequency9 andradio access technology. With this definition,LTE2600, LTE900, UMTS900, GSM900 are all dif-ferent layers.

Research papers on the intra frequency han-dover design and performance include [7] and[8], and many others.

3.2. Initial PLMN and Cell SelectionWhen powered on, the UE selects the PLMN

and the initial serving cell. In the processUE uses pre-stored infomation saved at previ-ous detach to speed up the initial cell selec-tion. After loss of coverage and in certain NASsignalling failure cases the UE makes a freshPLMN selection (without pre-stored informa-tion) in which different RATs are searched in aUE implementation specific order. Altenatively,the operator can influence the RAT search orderby storing radio access priorities on the USIM;

83GPP Rel8 defines only RSRP based threshold.9In case of GERAN the same frequency band.

8 J. Salo

0 20 40 60 80 100 120 140 160 180

−100

−95

−90

seconds

RS

RP

(dB

m)

COMPARISON OF RSRP, RSRQ, RS SINR, SSS SINR

0 20 40 60 80 100 120 140 160 180−20

−10

seconds

RS

RQ

(dB

)

0 20 40 60 80 100 120 140 160 180

0

10

20

seconds

RS

SIN

R (

dB)

0 20 40 60 80 100 120 140 160 180

0

10

20

seconds

SS

S S

INR

(dB

)

servingneighbour

both cells fully loaded only serving cell fully loaded

both cellsidle

Figure 3. Comparison of RSRP, RSRQ, RS SINR, S-SCH SINR for two sectors of the same BTS in fadingchannel conditions. Serving cell PCImod3=0, neighbour cell PCImod3=2. Both cells have two transmitantennas.


for example LTE may be searched prior to 3G.It should be noted that cell absolute priorities

are not used in initial cell selection.The full details of the initial PLMN and cell

selection procedure are described in more detailin 3GPP TS 36.304 and TS 23.122. In the follow-ing we assume that the PLMN and cell selectionhas been completed, and the UE is in the rese-lection phase.

3.3. Intra-Frequency Cell Reselection RulesA Rel8 UE may only camp on a cell that satis-

fies RSRP > RSRPmin. In turn, a 3GPP Rel9 UEmay only camp on a cell that satisfies both

RSRP > RSRPmin (10)

RSRQ > RSRQmin , (11)

assuming that UE transmit power is not lim-ited10 by cell broadcast settings.

If RSRQmin is not broadcasted, UE assumesRSRQmin = −∞, in other words the RSRQ cri-terion is always satisfied for all cells. Whilein Rel9 camping on a cell requires that bothRSRP and RSRQ criteria are fulfilled, the actualcell ranking in intra-frequency cell reselection isonly based on RSRP (among cells satisfying theabove criteria).

For equal priority cells, the basic form ofintra-frequency cell ranking criterion in 3GPPRel9 is

Rs = RSRPs − RSRPmin + Qhyst (12)

Rn = RSRPn − RSRPmin , (13)

where Rs and Rn are the R-criterion for serv-ing cell and nth neighbour cell, respectively.The quantities are average values given in deci-bel scale and the parameter RSRPmin (termed

10Otherwise an offset, equal to the difference of the maxi-mum UE transmit power capability and maximum allowedUE transmit power in the cell, is applied to the measuredRSRP value. As in 2G and 3G, the motivation of the offset isto prevent UE from attempting cell access when uplink linkbudget is limited.

Qrxlevmin in 3GPP specification) for both serv-ing and neighbour cells is broadcast on BCCH11.Hysteresis value Qhyst can be applied to servingcell RSRP measurement, in order to reduce ping-pong cell reselection. The cell ranking criterionis evaluated for all detected cells and the UE re-selects to the cell that has the highest R value.

In LTE, it is not mandatory to broadcastany neighbour cell information for cell rese-lection purposes. However, sometimes spe-cial neighbour-specific reselection offsets areneeded. For this purpose up to 16 cell individ-ual offset values can be broadcasted in BCCHSIB4, and UE ranks the cell, identified by its PCI,according to

Rn = RSRPn + Qoffset,n − RSRPmin . (14)

Finally, a Rel8 UE is not required to per-form intra-frequency idle mode measurementsif the RSRP of the serving cell is stronger than athreshold defined by the parameter SIntraSearch.In turn, for a Rel9 UE the rule requires thatboth RSRP and RSRQ are better than thresholdsSIntraSearchP and SIntraSearchQ, respectively. Thesearch thresholds are optionally transmitted onBCCH and in the absence of these parametersthe UE default behaviour is to search and mea-sure intra-frequency cells continuously regard-less of serving cell RSRP and RSRQ.

3.4. Intra-Frequency Cell Reselection Mea-surement Processing

The UE continuously scans for new neigh-bour cells and evaluates the cell reselection R-criterion for the cells it detects. The measure-ment requirement has been engineered in sucha way that the UE only needs to measure cellsat every paging opportunity of the serving cell,therefore maximizing battery savings. Eachneighbour cell is averaged over an integer mul-tiple of the idle mode DRX cycle defined in thetable below. A minimum of two RSRP/RSRQ11One Qrxlevmin value is broadcast for the serving cell in SIB1and another one for all intra-frequency neighbour cells inSIB3. Typically the same value of Qrxlevmin would be usedfor cells in a given macrocell layer.

10 J. Salo

measurement samples should be averaged forRSRP/RSRQ evaluation; hence the averagingtime is at least as long as the sampling inter-val. It can be seen that the worst-case samplinginterval in idle mode is 2.56 seconds, see Table2.

Table 2RSRP/RSRQ measurement sampling intervalfor intra-frequency cells, based on 3GPP TS36.133.

DRX cycle [sec] sampling interval [sec](number of DRX cycles)

0.32 1.28 (4)0.64 1.28 (2)1.28 1.28 (1)2.56 2.56 (1)

The UE is required to evaluate the R criterion,and possibly reselect to a new cell, at least onceevery DRX cycle. From planning and optimiza-tion point of view the cell reselection delay is ofsome interest. There are two subcases to con-sider: i) a cell that has not been detected hasbetter RSRP than the serving cell; ii) an alreadydetected and measured cell has R criterion bet-ter than the serving cell. The former case canhappen when UE comes out, for example, of anelevator or turns around a street corner. Thedifference here to the second case is that the UEcan only scan a limited number of PCIs during aDRX cycle, and hence detecting a new cell mayrequire several DRX cycles therefore increasingthe cell reselection delay. The cell reselectiontime for the two subcases is given in Table 3.For example, when idle mode DRX cycle is 0.64seconds selecting to an undetected cell may takeup to about 18 seconds, while for an already de-tected cell the time is about 5 seconds. Here ithas been assumed that Treselection = 0 seconds. Ifa positive value of Treselection timer is used, thenthe neighbour cell has to fulfill the R criterionfor at least Treselection seconds before cell reselec-tion can be made.

From Table 3 the difference between unde-tected and detected cell reselection time can beinterpreted as the time it takes for UE to scanover the entire PCI range; the scanning intervalover the range of 504 PCIs can be interpreted tobe 20 DRX cycles, in other words at least 25 PCIsshould be scanned during every DRX cycle.

Table 3Requirements for intra-frequency cell reselec-tion time when Treselection = 0, based on 3GPPTS 36.133.

DRX cy-cle length[s]

Reselection timeto undetectedcell [s]

Reselection timeto detected cell[s]

0.32 11.52 (36) 5.12 (16)0.64 17.92 (28) 5.12 (8)1.28 32 (25) 6.4 (5)2.56 58.88 (23) 7.68 (3)

4. INTRA-FREQUENCY CONNECTEDMODE MOBILITY

In this section measurement processing andbasic handover triggers in connected mode arediscussed. Again, the reader is reminded thatthe focus in this paper is on intra-frequency han-dovers.

4.1. Measurement ProcessingIn connected mode, the UE continuously

scans for new neighbour cells. Slightly simpli-fying the requirements in TS 36.133, a UE con-figured in connected mode DRX must be ableto detect a neighbour cell in less than 800 ms.The measurement processing described in thefollowing applies only to cells that have been de-tected by the UE. When long DRX cycle is in use,the time to detect a new cell is 20 . . . 40 DRX cy-cles. In the very worst case (2.56 second long cy-cle), the detection may take up to 20 · 2.56 = 51.2seconds. (In practice UE would probably moveto idle mode before this, or would transmitsome data and exit the DRX long cycle.) Hence


connected mode DRX is expected to degradehandover measurement performance in the net-work.

For each detected cell, the UE makes a phys-ical layer RSRP/RSRQ measurement, Mn, andprovides results to RRC layer for averaging onceevery 200 ms12. The physical layer measure-ment accuracy requirements were given in Table1. The raw measurement result Mn at time in-stant n is filtered at RRC layer (also called L3)using the formula

Fn = (1 − β)Fn−1 + βMn , (15)

where F0 = M0 and the filter coefficient k inβ = 2−k/4 is optionally signalled to UE in RRCmeasurement configuration message. The filter-ing is done in logarithmic domain. In case k isnot explicitly signalled, the default value k = 4,i.e., β = 0.5, is used. The filter coefficient β is de-fined separately for RSRP and RSRQ. The valueβ = 1 corresponds to the case with no L3 filter-ing.

4.2. Power Budget Handover Based on the A3Criterion

Perhaps the most basic measurement event isthe A3 event, or the "power budget" handovertrigger event. In its basic form the UE sends anA3 measurement report when a non-serving cellRSRP becomes better than the serving cell RSRPby a margin defined by an A3 offset parameter.In other words, when

∆RSRP > A3 offset, (16)

where ∆RSRP = RSRPneigh − RSRPserv. The∆RSRP is calculated for the L3 filtered valuesobtained from Eq. (15). The condition musthold for a duration defined by a time-to-triggerparameter. As with the 2G power budget han-dover, the motivation here is to keep UE con-nected to the cell that has the least downlink

12The value of 200ms is the nominal measurement periodfrom L3 point of view, UE may also perform measurementsmore frequently.

RSRPRSRQ

serving

neighbour

time

A3 offset

time to trigger

UE sends firstmeasurement report

UE stops sendingmeasurement reports

time to trigger

Figure 4. A3 measurement event criterion.RSRQ A3 measurement applies to interfre-quency HO only.

path loss13. Figure 4 shows an example of theA3 reporting event. During the time A3 condi-tion applies, the UE sends measurement reportsto BTS at reporting interval defined in the mea-surement configuration. Provided target cell se-lection and handover preparation to target cellsucceeds, the BTS will send a handover com-mand to the UE. The UE stops sending mea-surement reports when the A3 condition is nolonger fulfilled or a configured amount of mea-surement reports have been sent – or if the con-nection is released, handed over, or dropped. Ifdesired, a hystereris can be applied to activationand deactivation of the reporting condition butis neglected here for simplicity.

4.3. Coverage Handover Based on the A5 Cri-terion

The RSRP-based A5 measurement event is thesecond basic intra-frequency handover trigger14.

13Downlink path loss is defined as the difference betweenRS transmit power per subcarrier and RSRP, where RSRP, bydefinition, is equal or better than the RSRP of the stronger re-ceive antenna. Uplink path loss, however, depends on trans-mit antenna and thus may be higher than downlink pathloss if the power imbalance between UE antennas is large.14Using RSRQ for intra-frequency HO triggering is not nec-essary since RSSI is the same for source and target cell and

12 J. Salo

RSRPRSRQ

serving

neighbour

time

time to trigger

UE sends firstmeasurement report

UE stops sendingmeasurement reports

thr1

thr2

time to trigger

Figure 5. A5 measurement event criterion.

The UE sends a A5 measurement event whenRSRPserv < thr1 and RSRPneigh > thr2 for aduration defined by a time-to-trigger parame-ter, see Figure 5. The A5 event can be only betriggered when the RSRP from the serving isbelow threshold 1. The A5 event may becomehandy for triggering a fast handover in case ofrapid signal drop, with short time to trigger andsmall margin between the serving and neigh-bour cell. Using the A3 event to achieve a simi-lar fast reaction time, would result in increasedping-ponging in good coverage, which is not de-sirable. Another use case for the A5 event is loadbalancing between cell layers.

5. MOBILITY PARAMETER DESIGN

5.1. Design PrinciplesAny mobility parameter design should satisfy

some basic requirements, including:

• Handover margin should be small enoughso that no excessive interference is gener-ated by UEs connected to a cell with non-minimum path loss.

• The amount of unnecessary handovers

therefore the difference of RSRQs depends only on the dif-ference of RSRPs.

and ping-pong handovers should be mini-mized

• There should not be excessive handovertriggering delay which would increase calldrops when signal level drops sharply.

In addition, idle mode and connected modemobility parameters should be aligned so thatthere are no ping-pong cell changes in when UEchanges from idle to connected mode, and viceversa.

Obviously, other design criteria may be ap-plied, such as load balancing between cells.However, in this paper we focus on the simpleintra-layer use case where the goal is to set han-dover margin, time to trigger and L3 filtering co-efficient in such a way that the abovementionedcriteria are satisfied.

In general, the handover parameters shouldbe set differently depending on the service. Forexample, voice calls cause relatively low net-work load and hence a larger handover margincould be tolerated. On the other hand, voice calldrops due to late handover triggering should beminimized, as these cause noticeable end userdissatisfation.

5.2. Impact of A3 Offset on Uplink Interfer-ence

Figure 6 shows how increasing A3 offset re-sults in increased uplink interference in the net-work. Since not all UEs are connected to thecell with minimum path loss, they tend to useunnecessarily high transmit power thus creat-ing uplink interference. The simulation assump-tions for the result are listed in Table 4. Op-timization of uplink power control parameterscan reduce the SINR degradation, as can certainvendor-specific radio features such as inter-cellinterference coordination.

Downlink interference will not be impacted asmuch since there is no dynamic power controlfor the PDSCH channel. Therefore, the BTS usesa constant downlink transmit power per subcar-rierf; the transmit bandwidth in PRBs depends


−15 −10 −5 0 5 100

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

SINR, dB

CD

F

Mean number of UEs per TTI = 0.2

−15 −10 −5 0 5 100

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

SINR, dB

CD

FMean number of UEs per TTI = 2

0dB3dB6dB9dB

Figure 6. Network uplink SINR distribution fordifferent A3 offset values. Simulation assump-tions are given in Table 4.

however on vendor-specific link adaptation andscheduler implementation.

5.3. Impact of Time To Trigger and L3 FilteringCoefficient on Measurement Report Trig-gering Delay

Handover triggering delay is here defined asthe time difference between the time instantwhere the UE would send the measurement re-port assuming it had perfect RSRP/RSRQ in-formation available, and the actual time instantof sending the measurement report with non-ideal radio channel and measurement process-ing. The handover triggering delay is difficultto measure with real UEs, and hence to illus-trate the impact of parameters some simulationresults are shown in Figure 9. The simulation isbased on fading channel model at 2.6GHz withUE speed of 20 m/s. The scenario is shown inFigure 7, where it can be seen that at about 20-25 seconds after the beginning of the trace, theserving cell level experiences a transient (emu-lating a UE driving into a tunnel or being sud-

Table 4Simulation assumptions for Fig. 6.

Parameter ValuePath loss model Cost-Hata UrbanShadowing std 8 dBPenetration loss 20 dBBTS / UE antenna height 30 m / 1.5 mBTS / UE antenna gain 18 dBi / 0 dBiAntenna -3dB beamwidth 65 degreesAntenna downtilt 2 degreesNetwork layout 3GPP TS 25.914Number of sites / cells 19 / 57Inter-Site Distance 500 metersCarrier frequency 2.6 GHzSystem bandwidth 20 MHzBTS noise figure 2 dBUL power control P0 / α -100 dBm / 1Data model full buffer

denly shadowed by a building). A neighbourcell average RSRP is 5 dB less than the servingcell RSRP in the beginning of the trace prior tothe transition. The ideal "genie-aided" trigger-ing point would be at 24.5 seconds time instant,but due to signal fading of serving and neigh-bour cells as well as UE L3 filtering, there is cer-tain delay in triggering the handover; note thatin some cases fading may actually trigger a mea-surement report before the genie-aided point.Simulating a large number of fading realizationswith different values of A3 offset, time to triggerand β provides some insight into the impact ofthese parameters on the handover triggering de-lay.

Figure 8 illustrates how the L3 filtering in-fluences the measurement report delay for A3margin of 3dB and a few values of time to trig-ger. Firstly, if A3 margin and time to trigger arefixed, there is an optimum value of β that mini-mizes the delay. The second observation is thatthe delay is minimized for a fairly wide and con-sistent range of β = 0.25 . . . 0.5, almost indepen-dently of the time to trigger value. While thisbehaviour may seem illogical at first, the reason

14 J. Salo

5 10 15 20 25 30 35 40 45 50−30

−25

−20

−15

−10

−5

0

5Filtered RSRP for different values of β

time, seconds

L3 fi

ltere

d R

SR

P, d

B (

rela

tive)

β=1β=0.5β=0.11transientneighbour

Figure 7. Example of RSRP L3 filtering, UEspeed is 20 m/s, center frequency is 2.6GHz,and L3 sampling interval is 200ms.

0 0.2 0.4 0.6 0.8 10

2

4

6

8

10

12

beta

dela

y (s

econ

ds)

delay for A3 margin = 3dB

ttt=480msttt=1024msttt=2560msttt=5120ms

Figure 8. Impact of β on average handover mea-surement report delay for different values oftime to trigger. A3 offset is 3dB. UE speed = 20m/s, carrier frequency is 2.6GHz. Measurementbandwidth = 6PRB.

is that unfiltered fading (β = 1) in serving andneighbour cell can result in fast varying ∆RSRPvalues that frequenctly cross the A3 threshold,thus restarting the time to trigger timer. Mod-erate filtering β = 0.25 . . . 0.5 averages away thehigh signal peaks without overly increasing de-lay. Decreasing β to values below β ≈ 0.1 in-creases the filtering delay rapidly15.

The targeted delay bound can be met by sev-eral combinations of feasible parameter triplets(A3 offset, ttt, β). To ease design, another way ofplotting the inter-dependence of the parametersof interest is depicted in Figure 9. The Figure9 shows β-TTT parameter contours for A3 eventtriggering delays of 1, 2 and 4 seconds. The min-imum delay occurs at β = 0.25 . . . 0.5 for all eval-uated A3 margin values. If β is either near unityor very small, the delay is increased comparedto the minimum value. As an example, look-ing at the lower left subplot where A3 offset is3dB, one can see that if time-to-trigger is 2.56seconds, having either β ≈ 0.11 or β ≈ 0.8 bothresult in measurement report triggering delay offour seconds.

For design purposes, the radio planner firstchooses some maximum allowed measurementreport triggering delay, ideally, based on ser-vice type (real-time or non-real-time), carrierfrequency and UE speed. A design curve suchas the one in Figure 9 can then be used to definea range of allowed handover parameters. Mea-surement report triggering delay is usually notthe only design criterion for handover parame-ters; examples of multi-criteria design are pre-sented in other sections of this paper.

It should also be remembered that the totalend to end intra-frequency handover delay in-cludes:

• neighbour cell detection delay, 3GPP re-quirement is < 0.8sec when DRX is not inuse

• measurement report transmission delay,

15Filter coefficients k = 13 and k = 15 correspond to β =

2−134 ≈ 0.11 and β = 2−

154 ≈ 0.07, respectively.


beta

time to trigger (ms)

A3 margin = 2 dB

1 2

44

0 1024 2560 51200

0.25

0.50

0.75

1

beta


A3 margin = 1 dB

1 2 4

0 1024 2560 51200

0.25

0.50

0.75

1

beta


A3 margin = 3 dB

1 2

4

0 1024 2560 51200

0.25

0.50

0.75

1

beta


A3 margin = 4 dB

1 2 4

0 1024 2560 51200

0.25

0.50

0.75

1

Figure 9. Impact of β and time-to-trigger on handover measurement report delay. Delay contours for1, 2, 4 seconds shown for various values of A3 offset. UE speed = 20 m/s. Measurement bandwidth =6PRB.

16 J. Salo

including delays due to uplink schedul-ing request and measurement reportHARQ/RLC retransmissions

• BTS internal processing delay for measure-ment processing, handover triggering, tar-get cell selection, S1AP/X2AP signallingprocessing etc. These can be assumed tobe small, not more than a few millisec-onds.

• handover preparation delay due to X2 orS1 interface signalling, typically less than100ms, but depends on transport networktopology

• handover preparation processing at targetcell, for example radio and transport re-source reservation and admission control

• handover execution delay, consisting ofhandover command in source cell, cellsynchronization and random access delayin target cell, typically totalling about 20-30ms

• additional delays due to dedicated pream-ble retransmissions and RRC signalling re-transmissions at HARQ/RLC layer, in badradio conditions up to 500 ms

Therefore, for initial parameter design pur-poses, a margin of up to 2 seconds may beadded to the measurement report triggering de-lay curves. If connected mode discontinuous re-ception (DRX) is in use, the UE may choose towait until active time to send the measurementreport. This adds a random delay of up to theDRX cycle length to measurement report trig-gering.

5.4. Impact of Time to Trigger and A3 Offseton Measurement Report Triggering Delay

Figure 10 illustrates how time to trigger andA3 offset influence the measurement report trig-gering delay. Parameter contours correspondingto delays of 1, 2, 4 seconds are plotted for differ-ent values of L3 filtering coefficient, β. The samesimulation setup as in Figure 9 was used.

The figure again confirms the intuition thatincreasing A3 margin increases also measure-ment report triggering delay. As in Figure 9,several pairs of A3 offset and time to trigger val-ues result in the same triggering delay. The de-lay behaviour is quite insensitive to L3 filteringin value range β = 0.25 . . . 0.5 whereas valuesβ < 0.1 increase the delay strongly unless A3margin and time to trigger are also very small.

5.5. Impact of Time to Trigger and A3 Marginon Ping-Pong Probability

Figure 11 shows the impact of A3 margin andtime to trigger on the probability of UE beingunnecessarily handed over to a neighbour cell.The figure has been created using simulationwhere serving and neighbour cells have equalaverage RSRP, but undergo independent small-scale fading. The results are again shown forfour different values of β. The ping-pong prob-ability is the fraction of L3 filtered measurementsamples (200ms interval) that result in A3 eventtrigger.

It can be seen that as β becomes smaller theping-pong behaviour becomes less sensitive onthe value of time to trigger. It is interestingto notice that, to keep ping-pong probabilitysmall, increasing L3 filtering (decreasing β) re-quires increasing the time to trigger window.The reason is that with unfiltered fading theA3 criterion (difference of serving and neigh-bour RSRPs) signal frequently cris-crosses theA3 threshold which is why a long time to triggerefficiently mitigates ping-pong, as it is unlikelythat several consecutive A3 samples are abovethe threshold. On the other hand, L3 smoothingof serving and neighbour cell RSRPs results infewer A3 level crossings but also lengthens theduration of time the smoothed A3 value staysabove a certain threshold.

From handover parameter planning point ofview, the L3 filter coefficient is the same for allmeasurement events, while a dedicated time totrigger value can be defined for each event. Be-cause of this, one should choose the β value,so that it suits all handover purposes in the


A3

mar

gin

(dB

)


1 2

4

0 1024 2560 51200

1

2

3

4

5

6

A3

mar

gin

(dB

)


1

2

4

0 1024 2560 51200

1

2

3

4

5

6

A3

mar

gin

(dB

)


1

2

4

0 1024 2560 51200

1

2

3

4

5

6

A3

mar

gin

(dB

)


1

2

4

0 1024 2560 51200

1

2

3

4

5

6

β=0.5β=1

β=0.25 β=0.0743

Figure 10. Impact of A3 offset and time-to-trigger on handover measurement report delay. Delay con-tours for 1, 2, 4 seconds shown for various values of β. UE speed = 20 m/s. Measurement bandwidth= 6PRB.

18 J. Salo

network, and then tune the performance usingevent-specific time to trigger and power thresh-olds.

5.6. Other Handover ParametersIn addition to time to trigger, A3 margin and

L3 filtering coefficient, there are many otherhandover parameters that need to be config-ured. These are briefly summarized below.

• Reporting Interval. As long as the mea-surement event criterion is satisfied, theUE will keep sending periodic measure-ment report once every reporting interval.Reporting interval should be set jointlywith the filter coefficient; small β requiresusing longer reporting interval since lessnew information is introduced every mea-surement period (200ms). Further, set-ting too short a reporting interval not onlyconsumes more air interface capacity, butmore disturbingly it leaves less time forhandover preparation over the X2 or S1 in-terface as well as delivering the handovercommand in the source cell. Therefore set-ting reporting interval of non-urgent han-dovers (e.g. non-real time data) to lessthan 320ms is not recommended. Typi-cal value for A3 event reporting intervalis 320 . . . 1024ms.

• Hysteresis. Starting and stopping condi-tions for measurement reporting can bemodified with hysteresis. For example,assuming A3 margin of 3dB and hystere-sis of 1dB, the UE would start reportingwhen a neighbour cell is 4dB better thanthe serving cell, and stop reporting whenneighbour cell is less than 2dB better thanthe serving cell. In the earlier subsec-tions, it was assumed that hysteresis iszero dB, since usually there is no use casefor adopting non-zero hysteresis. Accord-ing to 3GPP TS 36.331 time to trigger ap-plies to both entry and exit condition forthe event reporting, therefore already in-cluding built-in hysteresis in time domain.

• Maximum number of reported cells. Maxi-mum number of reported cells can be con-figured, up to eight cells.

• Report Amount. The UE can be configuredto report a finite number of measurementreports (up to 64), or report indefinitely aslong as the measurement event conditionsare fulfilled. There is some benefit in con-figuring a finite amount of reports sinceotherwise, in case of missing or temporar-ily unavailable neighbour, a handover willnot be triggered and a stationary UE willcontinue to send measurement reports un-til it is moved to RRC idle.

• Report On Leave. This is a boolean parame-ter that determines whether the UE shouldsend one last measurement report at theinstant when the event reporting conditionstops applying.

UE can also be commanded to reportRSRP/RSRQ measurements periodically, in-stead of event-based reporting. This can be use-ful for optimization if the system is able to col-lect RF performance statistics from such reports,or use the periodic reporting data to automati-cally guide RRM algorithms.

6. PARAMETER DESIGN EXAMPLES

The purpose of this section is to illustrate thepresented concepts by means of simple parame-ter design examples.

6.1. Basic ExampleConsider a typical early LTE macrocell net-

work where majority of UEs are USB datamodems and the dominant service type is besteffort non-GBR data. The mobility in the net-work is low as UEs are mostly stationary. Thecarrier frequency is 2.6GHz. The followinggeneric requirements have been determined forthe initial mobility parameter design:

• SINR degradation (noise rise) should beminimized as much as possible



A3

mar

gin

(dB

)

0.0010.01

0.1

1024 2560 51200

1

2

3

4

5

6


A3

mar

gin

(dB

)

0.0010.010.1

1024 2560 51200

1

2

3

4

5

6


A3

mar

gin

(dB

)

0.0010.01

0.1

1024 2560 51200

1

2

3

4

5

6


A3

mar

gin

(dB

)

0.0010.01

0.1

1024 2560 51200

1

2

3

4

5

6

β=0.5β=1

β=0.25 β=0.0743

Figure 11. Ping-pong handover probability in the case of equally strong serving and neighbour cell,UE speed = 0.1 m/s, carrier frequency 2.6 GHz. A3 offset versus time to trigger for L3 filter values ofβ = {1, 0.5, 0.25, 0.0743}. Measurement bandwidth = 6PRB, SNR = 0dB.

20 J. Salo

• Total handover delay under quick powertransient should be less than 4 seconds

• Ping-pong handover probability at cellborder should be less than 1 percent, as-suming quasi-stationary UE.

Obviously, the conditions stated above aresomewhat contrived, for the sake of presentingthe following design example.

Only the A3 event based power budget han-dover is used in the network; the A5 coveragehandover will be added in the second part ofthe example. There are three main parametersto select: A3 margin, A3 time to trigger and theRSRP filtering coefficient β.

From Figure 6 it can be seen that A3 marginabove three decibels results in SINR CDF shiftof more than about 1 dB, hence it is preferrednot to have A3 margin higher than 3dB.

It is assumed that neighbour detection, HOpreparation and execution delay can be up totwo seconds; therefore the allowed measure-ment report triggering delay should not be morethan two seconds to keep the total handover de-lay less than four seconds. In Figures 9 and 10we hence focus on the parameter contours de-noting two seconds delay. From Figure 10 it isclear that at A3 margin values of at most 3dBand see that time to trigger values of 1024 msor less result in triggering delay of less than twoseconds, except when β = 0.0743. Selecting 3 dB/ 1024 ms as initial point, from lower left sub-plot of Figure 9 or from Figure 8 it is evidentthat all values in the range β = 0.2 . . . 1 resultin triggering delay of at most 2 seconds, indicat-ing that there is some freedom in selecting β. Tomake ping-pong handover probability as smallas possible, we select β = 0.25. The final han-dover parameters are:

• A3 margin of 3dB

• A3 time to trigger of 1024ms

• L3 filtering coefficient β of 0.25 (k = 8)

The idle mode reselection hysteresis, Qhyst isset to the same value as the A3 margin, to mini-mize unnecessary cell changes when UE movesfrom idle to connected mode, and vice versa.

Before moving forward, it should be notedthat there are other parameter combinations thatapproximately satisfy the boundary conditions.For example, using β = 0.5 instead would notsubstantially change the handover behaviour.

6.2. Basic Example ContinuedThe simple example from the previous sub-

section is continued. Assume that smartphonesare introduced to the network, and hence UEmobility increases due to subscribers browsingin trains, buses etc. The A3 handover works finein keeping the UE connected to the strongestcell, without excessive ping-ponging. However,in some cases UE received signal fades veryquickly down to RSRP levels below −120dBm,resulting in bad service quality and increasedprobability of call drop due to slow triggeringof the A3 event. If 2G/3G coverage is available,the UE can be directed to those safety layers(inter-layer HOs are not discussed further in thispaper). Prior to triggering inter-layer handover,it is worth trying to keep the UE on the servingLTE layer if a suitable cell can be found. For thispurpose, one can also utilize the A5 event withquick triggering. Assume that if serving cellRSRP < −120dBm, the UE would be directedto another layer. In this case, setting A5 thresh-old 1 (serving) and threshold 2 (neighbour) to,say, −118dBm and −117dBm with short timeto trigger of 160ms will make the last desper-ate attempt to move the UE to another intra-frequency cell, that now needs to be only 1dBstronger with short time to trigger. Should nosuch cell be available, the UE would then as alast resort be directed to another layer.

7. CONCLUSION

In this paper intra-frequency mobility param-eter design for LTE was discussed. Basic mea-surement quantities RSRP and RSRQ were re-


viewed and measurement example was shownto illustrate practical usage. The idle and con-nected mode mobility parameters were summa-rized. The intra-layer mobility parameter designproblem was defined in terms of three primaryhandover parameters: handover margin, time totrigger and measurement filter coefficient. In or-der to find a set of feasible solutions for theseparameters, the boundary conditions for uplinknoise rise, handover delay and ping-pong han-dover probability were considered. To illustratethe framework a simplified parameter planningexample was given.

Finally, as is well-known to practioners in thefield, radio network mobility parameters shouldnot be left to initially planned values, such asthose obtained based on the framework pre-sented here. Rather, constant optimization pro-cess should be in place to ensure sufficient per-formance as network load increases and newservices are introduced.

REFERENCES

1. H. Holma et al (eds.), LTE for UMTS, Wiley,2009.

2. F. Khan, LTE for 4G Mobile Broadband,Cambridge University Press, 2009.

3. S. Sesia et al (eds.), LTE, The UMTS LongTerm Evolution: From Theory to Practice,Wiley, 2009.

4. 3GPP TS 25.814, v7.1.0, 2006.5. 3GPP TS 36.133, v8.18.0, July 2012.6. 3GPP TS 36.133, v10.1.0, March 2011.7. P. Legg et al, A Simulation Study of LTE

Intra-Frequency Handover Performance,IEEE VTC, 2010

8. M. Anas et al, Performance Evaluation ofReceived Signal Strength Based Hard Han-dover for UTRAN LTE, IEEE VTC, 2007

Jari Salo received the degrees of Master of

Science in Technology and Doctor of Science inTechnology from Helsinki University of Tech-nology (TKK) in 2000 and 2006, respectively.Since 2006, he has been with European Com-munications Engineering Ltd, where he worksas a consultant for mobile wireless industry andoperators. His present interest is in radio andIP transmission planning and optimization forwireless networks. He has written or co-written14 IEE/IEEE journal papers, 35 conference pa-pers, and is the recipient of the Neal Shep-herd Memorial Best Propagation Paper Awardfrom the IEEE Vehicular Technology Society,for the best radio propagation paper publishedin IEEE Transactions on Vehicular Technologyduring 2007. He is also a co-creator of thewide-band spatial channel model adopted by3GPP for LTE system-level simulations. E-mail:[email protected]

mobility parameter planning for 3gpp lte: basic concepts and

Documents