Experiences of operating a pre-commercial WCDMA network

Introduction

During 2002, operators around the worldare planning to introduce third-generationmobile services to their subscribers. Manyhave selected the WCDMA system fromEricsson for building the infrastructure toprovide these services. The WCDMA stan-dard for the mobile network is the core tech-nology for efficient mobile multimedia ser-vices and for realizing the vision of true Mobile Internet access. The standard, which

has been developed after many years of re-search at Ericsson, is now the state-of-the-art telecommunications technology. Onearea of focus of the research has been to getadequate performance in the air interface(between the network and user equipment)in order to handle the new mobile data ser-vices such as Internet access, MMS, videocalls, and IP multimedia calls.

For commercial operations of third-generation mobile networks with payingcustomers, Ericsson is currently introducingWCDMA Release 2. Release 1 was intro-duced in 2001 for use as a pre-commercialservice by operators. It was used for the mainpart of the joint Ericsson-Vodafone pre-commercial tests. Release 1 has also beenused for demonstrating WCDMA features,network roll-out, network tuning, and forshowing compliance with technical re-quirements from regulators.

To ensure the expected performance at thecommercial introduction of the WCDMAsystem, Ericsson built a pre-commercialnetwork in cooperation with Vodafone UK.The network has been used to test variousoperating aspects in a realistic environmentbefore delivering hardware and software to

50 Ericsson Review No. 2, 2002

Experiences of operating a pre-commercialWCDMA networkAnders Birkedal, Eddie Corbett, Karim Jamal and Keith Woodfield

Ericsson and Vodafone have jointly built a pre-commercial WCDMA net-work in the UK. During 2001 and the first half of 2002, the network hasbeen used to evaluate various aspects of operating a WCDMA network,and for preparation for commercial services scheduled to start in 2002.The work mainly focused on evaluating the basic radio network character-istics of the WCDMA standard in a live environment.

The testing of a complete network under realistic operating conditionsbased on commercial hardware and pre-releases of software has beenextremely valuable to Ericsson and Vodafone.

The authors describe the network configuration, the test setup, variousareas of testing, and give some examples of test results.

2G Second-generation mobiletelecommunication systems

3G Third-generation mobiletelecommunication systems

3GPP Third-generation PartnershipProject

ALEX Active library explorerATM Asynchronous transfer modeBCCH Broadcast control channelBLER Block error rateBSC Base station controller (GSM

standard)BTS Base transceiver station (GSM

standard)CPP Cello packet platformCTR Cell traffic recordingDHCP Dynamic host configuration

protocolDNS Domain name serverEc/No Chip energy per noise spectral

densityFTP File transfer protocolGPEH General performance event

handlingGPRS General packet radio service

GPS Global positioning systemGSM Global system for mobile

communicationGSN GPRS support nodeIP Internet protocolISDN Integrated services digital networkISUP ISDN user partIu Interface between CN and RAN in

the WCDMA standardIub Interface between RNC and RBS in

the WCDMA standardIur Interface between RNC and another

RNC in the WCDMA standardMMS Multimedia messaging serviceMSC Mobile switching centerMub O&M interface for RBS (between

RBS and WCDMA OSS)Mur O&M interface for RNC (between

RBS and WCDMA OSS)NBAP Node-B application partNode-B Radio base station in 3GPP

standardNTP Network time protocolO&M Operation and maintenanceOSS Operations support system

PLMN Public land mobile networkPSTN Public switched telephone networkRAB Radio access bearerRAN Radio access networkRANAP RAN application partRBS Radio base station (in WCDMA

system)RNC Radio network controller (in

WCDMA system)RNS Radio network subsystemRNSAP RNS application partRRC Radio resource controlRSCP Received signal code powerSIR Signal-to-interference ratioTEMS Test Mobile System (product name)TU “Typical urban” channel modelUE User equipment (handset, terminal)UETR User equipment traffic recordingUu Interface between RAN and UE (air

interface) in the WCDMA standardVA Voice activityVAF VA factorVP Virtual pathWCDMA Wideband code-division multiple

access

BOX A, TERMS AND ABBREVIATIONS

Ericsson Review No. 2, 2002 51

operators. The core network parts of the sys-tem have been developed as an ongoing evo-lution from second-generation systems,whereas the radio access network representsa fundamental shift in technology, toachieve substantial gain in performance.Therefore, the activities associated with thepre-commercial network have focused onbasic radio characteristics that have a sig-nificant influence on overall system perfor-mance and represent a major technical chal-lenge for third-generation systems. Themain objectives have been • to evaluate the behavior of a complete

WCDMA system in a real-life environ-ment;

• to evaluate the performance and charac-teristics of the WCDMA radio access net-work (RAN); and

• to tune RAN system parameters under realistic radio conditions.

The network, which consists of 30 radio basestations (RBS) distributed to cover differenttypes of radio environments, was taken intooperation with WCDMA Release 1 softwarein April 2001, when the first live WCDMAcall was made. Release 2 of the software wasintroduced for testing in the spring of 2002.

Network configuration

GeneralThe pre-commercial network consists of 30radio base stations, two radio network con-trollers (RNC), a supporting core network,and service network nodes. Twenty-five ofthe RBSs, located in the Thames Valley re-gion west of London, have been arranged toprovide coverage over an area of approxi-mately 300 square kilometers. The coveragearea includes a good mix of rural, suburbanand dense urban environments. Motorwaycoverage is also available to perform drivetests at higher velocities. Apart from the 25RBSs, all other equipment is located atEricsson’s and Vodafone’s premises in Newbury, UK.

The network was split into two separateradio network subsystems (RNS) so that avariety of tests could be executed simultaneously. Adequate flexibility wasprovided to enable the RBSs to move be-tween RNSs, thus allowing each network togrow or shrink according to requirements.For some tests, the entire network of 25 fieldRBSs was parented to one RNC.

A third RNC was added to test the inter-face between RNCs (Iur).

The key objective was to evaluate Ericsson’s WCDMA radio access network ina live environment. The core network, trans-port network, and service network nodeswere installed to allow this evaluation totake place (Figure 1).

Radio base stationThe network is composed of 30 radio basestations configured with one carrier fre-quency per sector (Figure 2). Twenty-eightof the RBSs have three sectors (3x1). At twofield sites the RBSs have been configured fortwo sectors. The Ericsson RBS3202 is basedon the Cello packet platform (CPP), a high-performance ATM switch.1-3 The key func-tions of the RBS are• layer-1 termination of the air interface;• inner-loop power control;• softer handover combining and splitting;

and• operation and maintenance (O&M).

DHCP/DNS/FTP server

Service network

RNC3810

RNC 3810(used for lurtests)

RNC3810

E1

MSC/HLR/VLR

30 x E1

GGSN

Radio base stations(RBS3202)

RouterAXI540

O&M Router

Iu-c/Gr

Iu-u

Gn Gn

Grlu

Gi

MGW

Ethernet switch

GomGom

SGSN-W

BL/TTA x 4

Ethernet

Protocol analyzer

Iub

Iub

Iub

Iu

Iu

Iur

Iu

OSS server

OSS server

AXD301RXI820

Iu/Mur/Mub

Mur/Mub

Iu/Mur/Mub

Transmissionswitching site

Local labRBSs

Remote labRBSs

Internet FirewallDNS

Web

Email

GPS system clock

RANOSPC

EMASPC

UNIX W/S

UNIX W/S

UNIX W/S

Figure 1Block diagram of the pre-commercial network.

Radio network controllerThe radio network controller—RNC3810—is also based on CPP (Fig-ure 3). Early test cases were executed on asingle subrack configuration; during thetrial, however, the RNC was upgraded to atwo-subrack configuration to keep italigned with the majority of commercialroll-out projects. The RNC terminates theIub interface to the RBSs, the Iu interface tothe core network, and the Iur interface toother RNCs (to support inter-RNS hand-over). The key functions of the RNC are• layer-2 and the radio resource control

(RRC) part of layer-3 termination of theair interface;

• radio resource management, includingouter-loop power control and admissionand congestion control (capacity manage-ment);

• control of mobility, including soft hand-over;

• the provision of transparent bearer ser-vices between the core network and userequipment (UE);

• the management of the supply of bearerservices with traffic classes to the core net-work; and

• operation and maintenance.

ATM switch—AXD301/RXI820Each RBS is connected to the network con-trol center by a dedicated 2 Mbit/s (E1) leasedline. On the Iub interface, all traffic and sig-naling between the RNC and each RBS arecarried over a single ATM virtual path (VP).The AXD301 is used for switching virtualpaths from all RBSs onto one of two 155 Mbit/s (STM-1) connections to theRNCs (Iub grooming). Reconfiguration ofthese virtual-path cross-connects and ATMand IP configuration changes in the RNCand RBS allow the RBSs to be re-parentedbetween the two RNCs. All connectionsfrom the RNC to the core network via theIu interface are also connected via theAXD301. This means that Iu and Iub inter-faces for all nodes are available in theAXD301. The addition of extra virtual-pathcross-connects and STM-1 interfaces made itpossible to monitor these interfaces on a pro-tocol analyzer in an efficient manner. Dur-ing the project, the AXD301 was supple-mented with a CPP-based RXI820, whichtook over the function of Iub grooming.

Media gatewayThe media gateway (also based on CPP) pro-vides the interface between the RNC and

the core network. In the pre-commercialnetwork, the media gateway was used as anATM cross-connect. All Iu user plane andcontrol, and O&M interfaces of the RNC(Mur) and RBS (Mub) are passed to the mediagateway via a single connection from theRNC. The media gateway serves to split theIu traffic between the • mobile switching center (MSC), which

handles voice and circuit-switched traffic;and

• GPRS support node (GSN), which amongother things, handles packet-switchedtraffic in WCDMA networks.

The media gateway, which is a single sub-rack solution, also provides interfaces to theO&M router.

Core networkThe core network consists of• MSC;• GSNs;• home location register (HLR);• authentication register;• flexible numbering register;• equipment identity register; and• ISUP/ISDN protocol converter and

PSTN interface for testing 64 kbit/s circuit-switched data calls.

WCDMA OSSThe WCDMA operations support system(WCDMA OSS) is a set of software for han-dling operation and maintenance tasks forthe WCDMA radio access network. It givesa consolidated view of information on theradio access network, such as alarms, config-urations and basic performance. It also pro-vides several interfaces for easy integrationinto the existing management environment.

Operation and maintenanceinfrastructure

O&M router

The main purpose of the O&M router is toact as an interface between the IP-over-ATM-based WCDMA radio access networkand the IP-over-Ethernet-based O&M net-work, which consists of network servers andworkstations.

Network servers

A variety of UNIX and Linux-based serversoperate as• WCDMA OSS servers;• FTP/DNS/DHCP server;• NTP server, including GPS (for network

synchronization);


Figure 2The Ericsson RBS3202.

Figure 3The Ericsson RNC3810.


• network address translation (NAT);• Internet firewall;• Web server; and• active library explorer (ALEX) server—

for storing customer product information.

Workstations

A variety of UNIX, Linux and MicrosoftWindows-based workstations are used as• RBS and RNC element managers;• WCDMA OSS clients; and• UNIX workstations for command line

interface for data logging and debug toolsfor the RNC and RBS.

Test setup and toolsWhen testing traffic functionality, we gen-erally distinguish between • unloaded tests, where only a single UE is

used; and • load tests, where many UEs are used. The basic WCDMA traffic functions, suchas power control and soft handover are firstassessed in an unloaded situation. This en-sures that the functions work as expectedwhen the system is not under stress—thatis, when interference is largely stationaryand the main contributor is thermal noise.

In a loaded WCDMA system, interfer-ence, as seen by one radio link, consists ofthermal noise and other user signals. Inter-ference can thus vary considerably, in par-ticular when data users are few. Since theinner-loop power control mechanism inboth links is based on the signal-to-interference ratio (SIR), the power control

characteristics need to be evaluated duringload. The same is true for soft handover func-tionality, as well as the combination of softhandover and power control. Another areaof assessment is the basic behavior of the ca-pacity control functions (admission controland congestion control) of the WCDMAradio access network.

Generating real load on the air interfacefor controlled tests on a live pre-commercialWCDMA system involves considerable lo-gistical planning, since numerous UEs areneeded, and each UE must be controlled tosome degree. Data is logged extensively forone specific UE. The other UEs generate in-terference in a realistic way. Figure 4 showsa typical setup for a load test, where several“interferer” vehicles are used. Each interfer-er vehicle contains several UEs; the “test”vehicle contains the UE being used for datalogging. The interferer vehicles either holdstationary positions throughout the cell ordrive at random while the test vehicle typ-ically drives along a specific drive route. Thesame basic setup can be repeated for varioustypes of environment.

The capacity control functions are testedby adding calls in a cell using stationary ormoving UE clusters within the cell and ob-serving the network-side measuremententities used by the capacity control func-tions, and the actual behavior of the func-tions themselves. The unloaded tests are eas-ier to perform, since only the test vehicle isneeded.

To accommodate convenient UE-sidehandling and data logging, several cargo

Test vehicle

Cell 2

Cell 3

Cell 1

Interferer vehicle

In softhandover

Figure 4Setup for testing system load.

vans were fitted to include roof-mounted an-tennas, GPS receiver, uninterruptible powersupplies and a workbench environment(Figure 5). The UEs were pre-commercialprototypes of WCDMA terminals fromSony Ericsson and a variety of other UE sup-pliers. All logs on the UE side were collect-ed using laptop computers connected to the

UEs and running special logging software.The internal clock of each laptop computerwas synchronized to GPS time for subse-quent time correlation to network-side logs.

For enhanced observability, and to get asecond independent reading of the cell sta-tus in the field, pilot scanning tools wereused, including TEMS InvestigationWCDMA from Ericsson. Besides being usedas tools for evaluating the radio access net-work, the TEMS Investigation WCDMAand other tools, such as TEMS Cell Planner,were themselves evaluated during the livetests.

To increase the interference generatedduring testing, the interferer UEs used thebuilt-in antenna instead of the roof-mounted antenna—the use of built-in an-tennas increases path loss.

On the network side, the main observa-tion points were the RBSs, the RNCs, andthe Iub and Iu reference points. A protocolanalyzer was used to monitor the Iu, Iub andIur reference points (note also that Uu pro-tocols can be monitored by connecting a pro-tocol analyzer to Iub). The main protocols ofinterest were RRC (Uu), NBAP (Iub),RANAP (Iu) and RNSAP (Iur). In addition,the UE traffic recording (UETR) functionwas used for some entities. With these pro-cedures it was possible to log layer-3 mes-sages and to measure signal quality—for ex-ample, block error rate (BLER). All nodesin the network and the protocol analyzer ranan NTP client that was synchronized toNTP server software. This, in turn, was syn-chronized to GPS to ensure that the time-and-calendar function in each node was syn-chronized. The network logs could thus becorrelated with those collected on the UEside. Special parser software was developedto extract relevant entities from the networklogs. All resulting log files were saved to-gether with the UE log files for further pro-cessing and analysis.

In WCDMA, the source activity directlyinfluences how much power is being trans-mitted from the UE or RBS. Therefore, italso directly affects interference in the sys-tem. For this reason, to reproduce tests andto interpret the results properly, the sourceactivity for both the “test” and “interferer”UEs had to be controlled. Known sourcedata statistics as well as tools that generatereproducible data streams were used forpacket-data calls. For UE-to-fixed-linevoice calls, an announcement machine wasdialed. The announcement machine was fedby a voice recording with controlled voice


UE

Laptop

WCDMA scanner

GPS receiver

Figure 5Test equipment installation in the cargovans.

RSCP [dBm]

-60 to -40 (457)-70 to -60 (598)-80 to -70 (376)-90 to -80 (70)

Figure 6Measured RSCP throughout the drive route through an urban RBS cluster.


activity (VA) statistics that correspond totypical voice. In the uplink, a similar record-ing was fed to the UE microphone. This so-lution was also used at both ends for UE-to-UE voice calls.

Test objects

Radio network planningTools for planning and dimensioningWCDMA radio networks are included aspart of the TEMS family of products fromEricsson. To evaluate, tune and improve themodels used to describe the WCDMA radionetwork, early feedback from the pre-commercial network is essential. Measure-ments were made• to verify that the coverage for voice and

data services meets expectations as pre-dicted by the planning tools;

• to verify that areas where the mobile ter-minals are in soft or softer handover stateare in line with predictions;

• to evaluate performance and usability ofradio network planning tools;

• to create an understanding of the effect ofkey parameters on the radio network; and

• to investigate suitable measures of radionetwork performance.

To verify the network plan, we measured thepilot signals from the radio base stations.Although limited to a few common chan-nels, the pilot survey can give an indicationof the coverage, interference and soft hand-over that can be expected during real com-mercial operations.

We can verify the cell plan by studyingthe results of the pilot survey. We can alsoidentify potential interference situationsthrough the detection of many equallystrong pilot signals. Figure 6 shows a plotof the received signal strength from the radiobase stations for a part of the network. Fig-ure 7 shows the handover regions in the samearea using the same drive route. This testcase shows an urban cluster made up of 22active cells.

Operation and maintenanceAnother objective for testing was to assessthe basic O&M functionality and character-istics in a live network environment. Areasof particular relevance include the stabilityof the system, the efficiency of upgradingsoftware, and the general ability to observethe system in terms of performance man-agement and alarms.

Ericsson’s management solution provides

an integrated management view of the over-all network. The architecture structure con-tains embedded element management andsubnetwork management. The embeddedelement management is a fundamental prin-ciple in the O&M system. Each network el-ement (RNC or RBS) contains software forelement management—that is, the man-agement of that specific network element.

The main functions supported by the em-bedded element manager are• fault management—viewing and ac-

knowledging active alarms;• configuration management—configura-

tion of the hardware as well as settingradio and transport network parameters;and

• software management—upgrading thenetwork element with new software andtaking backup copies of the active con-figuration.

Ericsson’s subnetwork manager for theWCDMA radio access network (WCDMA

Active set size1 (1121)2 (316)3 (56)4 (4)5 (1)

Figure 7Number of legs in the active set throughout the drive route through an urban RBS cluster.The active set includes legs within 3 dB from the main signal.

OSS), supports the management of a singlenetwork element or coordinated manage-ment of multiple network elements. As withelement management, the main functionssupported by WCDMA OSS are fault man-agement, configuration management, andsoftware management. However the scopeof management is the entire radio networkincluding the transport network. In addi-tion, WCDMA OSS embodies several otherimportant features including product in-ventory functions and performance man-agement functions, such as performance sta-tistics and performance recording.

Stability and fault management

Typical tests of stability and fault manage-ment include Iub link breakage—for exam-ple, to simulate a microwave link in badweather. Another interesting type of test isto remove boards from nodes during opera-tion and when calls have been set up. Whenremoved, only directly affected trafficshould be lost and proper alarms should begenerated. After re-insertion, the boardsshould become fully operational. Otherareas include recovery testing, such as noderestarts and the accuracy and stability ofnode synchronization. This latter test is ofparticular interest in a network with vari-ous transmission media and configurations(for instance, optical links, microwave links,and cascading).

Software management

Another relevant test area is the loading ofnew software onto every node in the system.The Ericsson system will support softwareupgrades that do not affect traffic. The soft-ware of the RNC and RBSs can be remote-ly upgraded in parallel from WCDMA OSSby invoking the system upgrade function.This function automatically fetches newsoftware from a commonly accessible serv-er, transfers it to the node, and performs theactual upgrade. Another software-management function is backup adminis-tration, which allows for backup copies ofthe software to be made throughout the net-work and transferred to FTP servers. The software-management functions were eval-uated during use in routine day-to-day operation of the pre-commercial network.

Performance management

The Ericsson system allows the operator togenerate statistics (performance statisticsfunction) and make event recordings (per-formance recording) on UE or cells—UE

traffic recording (UETR) and cell trafficrecording (CTR). The operator may also setup events to trigger recordings or statisticscollection—general performance event han-dling (GPEH). The performance statisticsfunction collects samples over a given RNSfrom various operator-defined entities. Ex-amples include number of call setup failures,number of admission rejections, and theircauses. The results are collected in files thatmay be retrieved from the nodes throughWCDMA OSS. After having verified theperformance-management functions, theywere used as tools to test traffic functions.

Mobility functions

Cell selection/reselection

The cell selection/reselection function en-sures that the UE successfully camps on thecell that offers the best quality. Note thatthis procedure is used when the UE is in idlemode as well as when there is a packet-switched connection.

Given a predetermined PLMN, the UEuses a cell-selection procedure to find a cellon which to camp. Immediately after the UEhas powered on, the initial cell-selectionprocedure executes a cell-search procedure,which leads to the synchronization of fre-quency and time, and acquisition of thedownlink scrambling code of the selectedcell. On the other hand, when the UE is al-ready camping on a cell and has received aneighbor cell list from the network, the cellsearch procedure may be simplified, sincethe scrambling codes of the other cells arealready known.

From the camped normal state, the UEevaluates whether or not a new cell shouldbe selected. The measurements are al-lowed/triggered when • the quality (measured as Ec/No) drops

below a given threshold; • internal UE triggers are used; or • information pertaining to the broadcast

control channel (BCCH) is modified. When the measurements have been com-pleted, cells are ranked according to stan-dardized criteria and the UE reselects thecell with the highest ranking.

If the UE does not perform cell selectionand reselection properly, then it might se-lect a cell with high path loss. This couldresult in an unnecessary increase in interfer-ence in the cell, or a subsequent call set-upmight fail. Therefore, the main benefit oftesting cell selection/reselection came fromassessing



• the probability of camping on the best cellwhen in an environment that is “pollut-ed” with pilot signals; and

• the execution time of cell selection/re-selection, and parameter dependencies.

Soft and softer handover

During soft handover, the UE is simultane-ously connected to, is power-controlled by,and controls the power of, two or more cellsthat belong to different RBSs. The down-link and uplink signals are split and com-bined in the controlling RNC node. By con-trast, during softer hand-over the UE com-municates with two or more cells that be-long to the same RBS, and the downlink anduplink signals are split and combined in thatnode. The UE uses the same method forcombining downlink signals during softand softer handover. The set of cells withwhich the UE is communicating simulta-neously is called the active set.

In WCDMA systems, handovers are trig-gered by the UE in response to events ob-served among the cells it monitors. Whenthe quality of the downlink of a certain cellfulfills the stipulated criteria, the UE sendsa measurement report to the RNC, sug-gesting, as it were, that the active set shouldbe changed. The RNC then determineswhether or not the active set should bechanged.

The tests of soft and softer handover wereprimarily evaluated in terms of • voice quality—to verify that handovers

remain non-audible in various scenarios;• stability—to tune parameters, to give ap-

propriate levels of active set changes underdifferent conditions;

• handover delay—to assess the time afterthe radio conditions trigger a measure-ment report until a change has been madeto the active set;

• dropped calls—to assure call retainabili-ty at handover; and

• handover gain—to study the reduction intransmitting power obtained with softand softer handover.

Handover delay is a measure of the time thatelapses from the occurrence of an event inthe UE until the active set has been updat-ed. There are several components of thisdelay that can generally be grouped into• UE delay, which results from internal UE

processes;• network delay, which results from inter-

nal radio access network processing andsignaling; and

• UE network signaling delay, which re-

sults from communication between theradio access network and the UE.

Figure 8 shows the delay. Many componentsof the delay can be estimated by observingthe timing of messages sent between the UEand the network. The measurement controlmessage, which is sent after the active sethas been modified, contains an updated listof the cells that the UE is to monitor. TheUE cannot generate a new measurement re-port—and initiate a new handover—untilit has received this message and updated itsmeasurement processing accordingly.

The handover delay tests showed that thevarious components of delay were reasonable-sized contributions to the totalhandover delay, and that changes to the ac-tive set were typically completed within theexpected timeframe for the parameter set-tings used in the test.

Handover between third- and second-generation systems

An important function for the migrationfrom second- to third-generation systems ishandover between WCDMA and GSM. In

Network UE

UE-network signalingdelay

Network delays UE delays

Filtering and UE delay

UE processing delay

Network processingdelay

Network processing delay

measurementReport

measurementControl

activeSetUpdate

activeSetUpdateComplete

Figure 8Soft/softer handover: delays when updating the active set.

this context, the main characteristics to eval-uate are the execution time and end-userperception of handover in various radio-related scenarios. To test this function in thefield, a standard operational GSM radio net-work is used covering the same area as theWCDMA radio network. Some of the GSMBTSs are co-located with WCDMA RBSs.Various dual-mode 3G/2G UEs will be usedfor tests in this network when they becomeavailable from handset vendors.

Radio resource management

Power control

Since every cell uses the same frequency,each user’s signal interferes with the signalsof other users in the area. Therefore, powercontrol algorithms have a fundamental rolein WCDMA. The output power of both theUE and the RBS must be held to the mini-mum level needed to support each radio linkwith sufficient quality.

The power control of dedicated physicalchannels of the uplink and the downlinkwork in more or less the same way, and gen-erally constitute an inner-loop and outer-

loop algorithm. The task of the outer loopis to keep a measured BLER level at a spec-ified BLER target level. This is accom-plished by continuously updating an SIRtarget level, which in turn is used by theinner loop. With the given SIR target levelas input, the inner loop strives to maintaina measured SIR level at the SIR target levelby requesting that the transmitting sideshould increase or decrease its output powerlevel.

Thus, by using the outer loop, the con-nection quality is maintained in terms ofkeeping the BLER at a predetermined value,rather than keeping the SIR constant. Thisis relevant since the BLER versus SIR rela-tionship differs for different radio environ-ments and UE speeds.

For uplink power control, the inner loopis managed by the RBS to control the powerof the UE. The outer loop is managed by theRNC—that is, the algorithm works on theBLER measured after soft handover com-bining. The downlink power control loopsare managed by the UE, where the inner loopcontrols the power of the RBS. One objec-tive of the tests was to evaluate the perfor-


[dB]7.0

6.5

6.0

5.5

5.0

4.5

4.0

3.550 10 15 20 25 30 35 40 45 50

Time [s]

SIRSIR target

Figure 9Typical time fragment of uplink SIR targetand SIR for an urban environment, 50 km/h, with 100% VAF.


mance of the downlink inner-loop powercontrol by plotting BLER versus the SIR tar-get and recording the difference between theSIR target and measured SIR for differentenvironments. We monitored the standarddeviation of the measured SIR and BLER andcompared it to the test during which down-link outer-loop power control was activated.

Another objective of the test was to eval-uate the performance of the downlink inner-and outer-loop power control. From thesetests we were able to determine the neces-sary SIR and SIR target, and standard devi-ation of measured SIR and BLER for differ-ent environments.

The performance of the uplink inner-looppower control was also evaluated. We wereable to characterize BLER versus the SIR tar-get, and to observe the difference betweenthe SIR target and measured SIR for differ-

ent environments. We monitored the stan-dard deviation of the measured SIR andBLER and compared it to the test duringwhich the uplink outer-loop power controlwas activated.

Similarly, we evaluated the performanceof the uplink inner- and outer-loop powercontrol. From these tests we could determinethe needed SIR and SIR target, and the stan-dard deviation of measured SIR and BLER.

The uplink tests illustrated in Figures 9and 10 were performed with the uplinkouter-loop algorithm activated with a BLERtarget of 0.5% and while driving in an urbanarea at 50 km/h with no antenna in line ofsight using a single UE. The variation inpath loss along the drive route was estimat-ed to be less than 20 dB. The average SIRtarget in this case was around 5 dB. For lowerspeeds, slightly lower values were measured,

Figure 10Distribution of SIR target, SIR and SIRerror for an urban environment, 50 km/h,with 100% VAF.

[%]

20

15

10

5

010 2 3 4 5 6 7 8 9 10

SIR-target [dB]

[%]

15

10

5

010 2 3 4 5 6 7 8 9 10

SIR [dB]

[%]

20

15

10

5

0–1.5–2 –1 –0.54 0 0.5 1 1.5 2

SIR error [dB]

as expected. A highway drive test gave anSIR target of as low as 4 dB, but since thedrive route was in line of sight at nearly alltimes, this was expected. In general, SIR fol-lows the SIR target, with a reasonable stan-dard deviation.

The measured BLER was generally closeto the target of 0.5%.

Admission and congestion control

Admission and congestion control are algo-rithms at the cell level that prevent and re-solve overload situations on the air interface.The admission and congestion control algo-rithms control load in the radio access net-work. They maximize capacity in the radioaccess network while maintaining the re-quested quality of service and coverage—that is, they guarantee stability in the sys-tem. The core network determines the qual-ity of service parameters for each radio ac-cess bearer (RAB). The admission and con-gestion control algorithms control the loadin the radio access network by regulatingthe use of resources.

The admission and congestion control al-gorithms regard air-interface resources aslimiting resources—that is, they mightlimit performance. Downlink power, uplinkinterference and downlink channelizationcodes need to be controlled. The resourcesare controlled by measuring resource load.The uplink resources are mainly controlledthrough a virtual resource—the air interfacespeech equivalent. Which resources mightbe limiting performance depends on the con-ditions of the specific cell, including systemdeployment, and operator control. The fol-lowing aspects of the admission and conges-tion control algorithms are especially inter-esting in terms of evaluation:• system stability—is the system stable rel-

ative to total downlink power usage anduplink noise rise?

• system robustness—is the system able torecover from overload situations?

• system capacity—what is the average ca-pacity provided in a cell? and

• connection quality—the quality of indi-vidual connections must be maintained.


Microsoft is a registered trademark ofMicrosoft Corporation.UNIX is a registered trademark of the OpenGroup.

TRADEMARKS


Three types of test case were specified for theperformed tests:1. Admission control in a single cell with a

static radio environment—to evaluate theadmission control with focus on the ad-mission process.

2.Admission and congestion control in asingle cell with a dynamic radio environ-ment—to evaluate congestion controlwith focus on the congestion process.

3.Admission and congestion control inmultiple cells with a dynamic radio envi-ronment—to evaluate the mobility set-tings of admission and congestion control.

ConclusionThe testing described in this article is asample of the WCDMA system evaluationconducted by Ericsson in the UK, in orderto prepare WCDMA technology in gener-al, and the air interface in particular, forfull-scale commercial operations. Theevaluation can be seen as the last link in along chain of verification made at differ-

ent levels of the WCDMA system to en-sure that it is ready for release to payingsubscribers.

The evaluation has shown that simula-tions cannot fully replace full-scale, real-lifetesting of complicated technology to verifysystem functionality and interoperabilitybetween its various parts. This exercise hasconsiderably decreased the probability ofunforeseen problems when commercial sys-tems are taken into operation. Moreover, ithas given Ericsson the opportunity to tunethe default system parameters stipulated inthe operating guidelines. This will save op-erators valuable time when they take the sys-tem into operation.

Finally, Ericsson and Vodafone eachgained valuable know-how concerningWCDMA radio characteristics and systembehavior. This can be applied to improve• radio network planning tools;• parameter-setting guidelines;• system performance in upgraded system

releases; and• skills in operating the network.

The authors want to thank the staff fromVodafone who participated in this project,and without whose contributions this articlewould not be possible.

ACKNOWLEDGEMENTS

1 Reinius, J.: Cello—An ATM transport andcontrol platform. Ericsson Review Vol.76(1999):2, pp.48-55.

2 Berglund; b:, Nygren, T. and Sahlman, K.-G.: RF multicarrier amplifier for third-generation systems. Ericsson Review Vol.78(2001):4, pp. 184-189.

3 Kling, L.-Ö., Lindholm, Å., Marklund, L.and Nilsson, G. B.: CPP—Cello packetplatform. Ericsson Review Vol. 79(2002):2,pp. 68-75.

REFERENCES

Experiences of operating a pre-commercial WCDMA network

Documents

Transcript of Experiences of operating a pre-commercial WCDMA network