third design release of ericsson’s wcdma macro radio … phases background ericsson’s strategy...

Initial phases

Background

Ericsson’s strategy for 3G network develop-ment is to release products and features inaccordance with customer needs at differentphases of network roll-out (Figure 1). Themost important customer needs during thefirst phase, commercial launch, are• rapid rollout (mainstream site concept);• efficient training of staff;• stability; and • future-proof investment (backward and

forward compatibility).

First RBS design release – commercial

launch

Ericsson released its first indoor and outdoormacro base stations (RBS 3202 and RBS 3101) in early 2001. The products werebased on the first commercial RBS designrelease (RBS R1).

Although time to market was a priority

consideration for this first release, Ericssondid not compromise on the future-proofnessof its products. For example, they could beexpanded to support six sectors and four car-riers with transmit and receiver diversity.

The commercially available technology in2001 – signal processing with digital sig-nal processors (DSP), field programmablegate arrays (FPGA), and linearization tech-nologies based on feed-forward techniques –enabled a 3x2 configuration in a single cab-inet with the same footprint as the GSMRBS 2000 base station (600x400mm). Thearchitecture also supported larger configu-rations via digital interconnection of up tofour cabinets.

It is worth noting that close collaborationbetween design, research and standardiza-tion projects ensured that the early archi-tecture would later also support HSDPAand E-UL. Indeed, provisions for the com-plete radio frequency (RF) transmitter chainwere built into the RBS R1 from the start,supporting the full HSDPA implementa-tion with higher-order modulation withoutdegradation of output power.

One other crucial design choice was tobase the RBS on the Connectivity PacketPlatform (CPP) for control, operation andmaintenance (O&M), transport network in-terfaces, switching functionality, and syn-chronization.1 CPP, the foundation of everyswitching node in Ericsson’s WCDMAradio access network, simplifies RAN oper-ation and maintenance, and in terms oftransmission functionality, it guaranteescommon evolution of the entire radio accessnetwork. Most importantly, however, CPPensures that every node in the radio accessnetwork is truly able to meet future re-quirements relative to evolving radio net-work functions and capacity (including IPtransport network and capacity enhance-ments such as HSDPA and E-UL).

Second RBS design release –

increasing capacity and performance

The requirements of the second phase,whose focus is on broad (nationwide) cover-age, can be summarized as follows:• greater flexibility (more configurations);• continued focus on outdoor macro cover-

age solutions; and • greater emphasis on in-building coverage

solutions.Ericsson’s second RBS design release (RBSR2) improved on the architecture and sub-systems in RBS R1 and introduced softwaresupport for additional configurations.

70 Ericsson Review No. 2, 2005

Third design release of Ericsson’s WCDMA macroradio base stations

Bo Berglund, Michael Englund and Jonas Lundstedt

The market for WCDMA has taken off in several regions around the world.

Europe, East Asia and Australia, for example, are each reporting acceler-

ated growth in subscriber uptake. Many operators, after a successful roll-

out of coverage, are now also offering high-quality networks that carry

steadily increasing loads of voice and data traffic.

Three 3G standards are competing for subscribers: WCDMA,

CDMA2000 and TD-SCDMA. To sustain continued growth in regions

where customers are accustomed to excellent 2G handsets, services and

high-speed fixed broadband access, operators of 3G networks must offer

even better services and greater mobility. Moreover, they must keep their

tarrifs competitive. Consequently operators are very interested in peak

performance, capacity, and cost-effectiveness.

This article discusses how Ericsson’s new, third release of its WCDMA

macro radio base stations (RBS) capitalize on advances in technology to

improve the architecture of the RBS node to meet the challenges

described above and to help operators target new business opportunities.

The new design enables operators to double node capacity, increase cov-

erage, simplify maintenance, and dramatically reduce power consump-

tion. The combined effect of these enhancements yields considerably

lower capital expenditures (CAPEX) and operating expenses (OPEX) in the

radio access network (RAN).

The authors briefly review Ericsson’s WCDMA RBS development strate-

gy, giving examples of important design choices and explaining how the

architecture has evolved to fit new market requirements and exploit

advances in technology. In particular, they discuss the improved RBS

architecture, advances in multicarrier power amplifier (MCPA) linearization

technology, and design aspects of importance to high-speed downlink

packet access (HSDPA) and the enhanced uplink (E-UL). The authors also

introduce Ericsson’s newest macro base station members of the RBS

3000 family.

Review nr 2 2005.qxd 05-12-16 15.07 Sida 70

Ericsson Review No. 2, 2005 71

Ericsson also broadened its RBS portfolio byintroducing the Super-compact (RBS3104), Micro (RBS 3303) and Main-remote(RBS 3402) base stations for challengingsites and dedicated in-building coverage so-lutions (Figure 2).

New ASICs enabled an eight-fold increasein uplink capacity per board. Likewise, newclipping algorithms and an enhanced feed-forward multicarrier power amplifier(MCPA) improved power efficiency from 9 to 11% and enabled a new, high-power30W power class of products.

New receiver algorithms and an improvedreceiver noise figure in the uplink (UL) ac-companied the new high-output power op-tion in the downlink (DL). These enhance-ments improved receiver sensitivity by upto 2dB and the effective coverage by morethan 20%.

As demonstrated by the in-service perfor-mance (ISP) figures collected from the morethan 40 commercially launched WCDMAnetworks powered by Ericsson nodes in theradio access network, the first and secondRBS design releases have each performedsatisfactorily with respect to set require-ments. Cell availability and dropped-callrates have continuously improved in recentyears, despite the addition of new RANfunctionality and increasing traffic. The av-erage ISP figures from these commerciallydeployed networks are already on par withthe ISP figures of mature GSM networks.

RBS R3 development

Changing needs, technology advances,

and lessons learned

Since the first launch of WCDMA, two as-pects in particular have changed: • OPEX has very rapidly become a prima-

ry operator concern (much more quicklythan in previous system generations suchas GSM); and

• the emergence of, and need for, new fre-quency bands, including the requirementfor dual-band implementations.

Slow uptake of traffic in 3G networks from2001-2004 put many operators in a finan-cial bind, forcing them to find ways to cuttheir operating and capital expenditures.Nearly 15% of an operator’s total costs canbe attributed to radio network-relatedOPEX; 10% to radio network-relatedCAPEX. In other words, the radio networkaccounts for nearly 25% of an operator’stotal costs. Therefore, in the context of cost

Figure 1Various phases of network rollout.

Figure 2Ericsson’s RBS hardware releases.


effectiveness, radio network-related OPEXand CAPEX play a very prominent role.

The network cost structure, briefly de-scribed in GSM network solutions for new-growth markets, consists of numerous com-ponents but it can be assumed that the totalcost is more or less proportional to the num-ber of sites in the network.2 The cost of theRBS, in turn, typically represents 20-40%of the cost of the site it occupies. Therefore,RBS enhancements that reduce the numberof sites as well as lower site cost have a largeimpact on overall cost. Examples of charac-teristics that affect number of sites are out-put power and receiver sensitivity. Exam-ples of characteristics that have a direct im-pact on site cost are:• RBS footprint (and required number of

cabinets); • power supply (flexibility, integrated or

external); and • cooling requirements (RBS power effi-

ciency). Power consumption and ease of installationand maintenance also directly affect cost.

Advancements in technology have madeit possible to increase capacity per node andRBS power efficiency. In particular, recentadvances in the areas of DSP capacity, base-band ASIC development, and MCPA lin-earization have vastly increased the poten-tial density or ratio of capacity per volume.Greater density means that one can grow ca-pacity without increasing footprint or vol-ume. Obviously, this facilitates site acqui-sition and planning. Improved power effi-ciency also reduces power bills and lowersCAPEX and OPEX associated with cooling,

power supply and battery backup systems. Experience of GSM and the two previous

WCDMA RBS design releases has proventhe importance of forward and backwardcompatibility. Operators have come to ex-pect that the functionality of installedequipment can be upgraded in harmonywith the rest of the network. In short, theRBS must represent a secure investment. Itmust be reliable, expandable and compati-ble with future investments. In this context,Ericsson made an important choice basingall its RAN nodes around a common 3Gplatform (CPP). The same can be said forEricsson’s unwavering stance on upgrade-ability and compatibility. Certainly thesechoices called for a fundamentally larger de-sign effort leading up to the first commer-cial release of the 3G RAN products, butbecause each subsequent 3G node fromEricsson derives from the same platform, op-erators know they can evolve their nodes.

The platform concept is one reason Ericsson has the most complete WCDMAproduct portfolio in the industry. It alreadycovers three frequency bands, and at leasttwo more frequency bands are planned forthe coming year.

Third RBS design release

Ericsson began its studies for RBS design re-lease 3 (RBS R3) in 2001. This was the sameyear that products based on RBS R1 ap-peared in the market. The objective of thestudies was to incorporate 3G RBS designexperience and experience gained from de-veloping and adapting GSM to new marketrequirements.


• 15-35% increase in nominal output power: this translates into increased downlink coverage

(up to 20%) and nominal output power of up to 120W per sector

• Improved static sensitivity: -128.5dBm for 3GPP 12.2kbps 10-3BER and two-way receiver

diversity (2100MHz)

• Twice as much capacity per cabinet: now up to 3x4 or 6x2 in one cabinet with Ericsson’s stan-

dard 600x400mm footprint with no requirements for ventilation space at sides or rear

• Dual-band support in one cabinet, for instance, 3x2 1900MHz + 3x2 850MHz

• Integrated RRU support: the reuse of existing site power, baseband and transmission infra-

structure yields low site cost during network expansion

• Reduced power consumption: the power consumption of a typical 3x2 20W (2100MHz) is

down 55% to 0.8kW

• 70% increase in availability, which translates into fewer site visits and lower repair costs

• Integrated AC and DC power options: a single-cabinet self-contained RBS can now support up

to 12 cells in two frequency bands

• Simplified architecture: compared to earlier design releases, a 3x1 configuration in R3 requires

half as many plug-in units (PIU), one-third as many O&M processors, and half as many inter-

connects

TABLE 1: SUMMARY OF DESIGN ENHANCEMENTS IN RBS R3.



The first two RBS design releases laid thefoundation for the O&M and transmissionplatform, radio performance, in-service per-formance, and upgradeability. The focus ofthe RBS R3 design effort was thus primar-ily on reducing total cost of ownership(TCO) and on facilitating greater R&D ef-fectiveness in preparation for the upcomingmultitude of frequency bands and configu-ration requirements. The focus, in terms ofreducing TCO, was on• increased-coverage solutions – which

translates into fewer sites;• more than twice the capacity per cabinet –

based on the same best-in-market foot-print;

• multiband support;• significantly lower power consumption;• even greater reliability through simplified

architecture and integration;• integrated power supply; and• drastically simplified maintenance

through modularity.The R&D efficiency was greatly enhancedthanks to reduced complexity in the archi-tecture and improved product flexibility, forexample, by developing more integratedsubsystems.

In addition, one of Ericsson’s strategic en-vironmental goals has been to reduce powerconsumption. An important conclusion ofextensive life-cycle assessments (LCA) con-ducted since the mid-1990s is that the mostsignificant impact of telecommunicationssystems on the environment is linked to en-ergy consumption from operations. Fur-thermore, in this context, radio base stationsare the single largest consumers of energy.

The RBS R3 development successfullytransitioned to series production in the firstquarter of 2005. Table 1 contains a briefsummary of the RBS R3 design enhance-ments.

RBS R3 architecture

Ericsson Review reported on RBS R1 andR2 products and architecture in 2000 and2003.3-4 Figure 3 shows how Ericsson im-proved modularity, going from R2 to R3 bymeans of higher-order integration. Inessence, Ericsson’s designers increased sub-system integration in virtually every RBSfunction area while maintaining compati-bility with important interfaces, such as Iuband Uu, antenna systems, and the internalbaseband. To operators, this means largerconfigurations in one cabinet, significantlylower power consumption and improved

availability – due to new MCPA lineariza-tion techniques, O&M and transport net-work integration, and an improved coolingconcept.

Flexible configurations and

compatibility

The market has been very clear in spellingout its demands for one-cabinet solutions.Many operators, for example, want self-contained indoor cabinet solutions with in-tegrated power. Likewise, dual-bandWCDMA is being built out in North Amer-ica and will also probably become a re-quirement in other markets. Certainly, aone-cabinet dual-band solution offers sever-al advantages: resource pooling, small foot-print, low power consumption, simpler ex-pansion and reduced maintenance. Ericssondesigned the new RBS R3 with these at-tributes in mind. It supports all plannedconfigurations in one cabinet, accommo-dating up to four carriers per sector, six-sector or dual-band configurations.

The new baseband boards (TXB andRAXB) are fully compatible with earlier releases, as are the new CPP boards, the exchange terminal boards (ETB) for trans-mission, and the general-purpose boards(GPB). The entire radio network can thusevolve very smoothly in terms of transport(for example, IP transmission) and radiofunctionality (such as HSDPA services).

Radio

Given the requirement to double the con-figuration capacity in one cabinet, the log-ical design objective for RBS R3 was to dou-

Figure 3

Comparative overview of the architecture, RBS R1 through RBS R3.


ble the available output power per cabinet.In fact, Ericsson exceeded this goal to im-prove individual carrier capacity.

Greater output power is always beneficialfor downlink coverage and capacity. The ex-tent of this benefit is determined by sever-al variables, such as cell size, the inclusionor exclusion of tower mounted amplifiersand antenna system controllers (TMA/ASC), feeder losses, and the mobile phone.In the 20W power class, for example, as-suming ASC and 3dB feeder loss, networktests have shown a close relationship be-tween “useful” output power and downlinkcell capacity. In other words, an increase to30W typically yields a 30% increase indownlink cell capacity.

The task of doubling available outputpower per cabinet (without increasing cab-inet size) entailed re-examining the entiredesign: power supply, cooling, integrationof subsystems, internal jumpers, filter de-sign, and power amplifiers.

Radio architectures with the greatest po-tential (in terms of efficiency, flexibility andcapacity per volume) make use of radio units(RU) and filter units (FU), where a radio unit is a complete transceiver and MCPA (Figure 3). The filter unit is composed of thefront end: transmitter and receiver cavity fil-ters, low-noise amplifiers, logic and inter-nal bias-Tee for communication, power feedover feeder to ASCs and remote electrical an-tenna tilt (RET), and lightning protection.

This simple and elegant architecture hasfew building blocks and few interfaces. Fur-thermore, it accommodates ongoing tech-nical evolution relative to power amplifierdesign, because the digital and analog partsof the transceiver are integrated with thepower amplifier.

The improved power consumption effi-ciency and maximum output power capa-bility are good examples of the achievementsof the new design. For example, the 3x220W configuration in the 2100MHz bandis more than twice as efficient as that of RBS R2. The maximum output power ismore than 400W for an R3 cabinet (at topof cabinet) as compared to approximately180W for the R2 cabinet. This large in-crease in output power capability facilitateslarge single and dual-band configurationswithout compromising downlink cell ca-pacity.

Baseband

Ericsson designed the RBS 3000 basebandarchitecture to ensure smooth and long-

term evolution of functionality and capaci-ty. One important feature of the basebandarchitecture is the separation of the uplinkand downlink into different resource pools.A drawback of this design choice is the needfor additional inter-board interfaces andthus greater need for architectural systemplanning, to ensure future compatibility.But once these hurdles had been cleared thebenefits were quite substantial. For in-stance, one can optimize the uplink anddownlink independently using differenttechnologies for each. One may also dimen-sion node capacity according to traffic needs,which improves cost-effectiveness. Thisbenefit will be especially pronounced as datatraffic volumes increase, because the down-link typically carries six times more datatraffic than the uplink. The design also dou-bles pooling efficiency by introducing larg-er resource trunks, and giving the systemfull freedom to use all the available resourceson all individual uplink and downlinkboards – that is, there are no restrictions puton the allocation of necessary downlink anduplink radio link resources as would havebeen the case had the resources been on thesame board.

Key characteristics of the baseband archi-tecture, to accommodate new functionalityand greater capacity throughout the lifetimeof 3G, are efficient resource utilization andhigh capacity. Although the channel ele-ment (CE) is a resource equivalent not stan-dardized by 3GPP and thus defined differ-ently by different vendors (the definitiondiffers in how many CE are required for agiven service, whether CE resources are re-quired for common signaling, compressedmode measurements, and so on), it repre-sents a simple and intuitive measurement ofbaseband capacity.

The RBS R3 architecture can boast thelargest baseband capacity in the industry ina single, standard-sized cabinet: 1536CE inboth the uplink and downlink. Given futureemphasis on downlink data services, thechannel element data efficiency is particular-ly high in the downlink. Ericsson’s 1536CEdata capacity in the downlink is equivalentto 2000-3100CE, depending on the industrynorm used to express number of channel elements for different data services.

Ericsson has employed higher-order inte-gration to obtain very high baseband capac-ity. Its most recent baseband boards (RAXBand TXB) use proprietary ASIC design andnew, high-capability DSPs to give 128 and384CE per board, respectively.




For HSDPA and E-UL, the baseband ar-chitecture employs large-scale pooling ofhigh-speed data resources and a commonscheduler. The present TX board supportsup to 45 HSDPA codes. By pooling theHSDPA downlink resources with R99downlink resources it is possible to optimizethe scheduler in terms of available downlinkpower and traffic. A fast scheduler has a pos-itive impact on network latency, or in otherwords, on the end-user experience.

The RBS R3 architecture maintains com-patible internal baseband interfaces and em-ploys high-capacity boards to serve higher-order configurations. Every uplink anddownlink board is compatible with RBS R1and R2, which is to say the entire networkcan benefit from functional and performanceenhancements to the baseband.

Control

The RBS control architecture, based onCPP, provides switching and basic O&Mand transport functionality.1 The mainprocessor handles RBS operation and main-tenance and controls traffic and switching.Processing power can be pooled via severalmain processors, for example, for redundan-cy, to handle increased traffic, or to providegreater ATM adaptation layer-2 (AAL2)switching capacity.

The control base unit (CBU) is a new con-trol subsystem that integrates a main proces-sor, 16Gbps switch core, timing unit, andE1/T1/J1 interface board. It has been intro-duced to improve space efficiency, power ef-ficiency, and increase availability. The CBUrepresents a minimum set of mandatoryfunctions for any configuration. The com-bined subsystem is half as large as the corre-sponding size of the subsystems it replaces.

Transport

The transport functionality is based on CPP.Different exchange terminal boards providethe optical or electrical interfaces to SDH,PDH, ATM or IP networks. An internalswitch core handles dedicated point-to-point connections between the ETB and thecorresponding RAXB, TXB or anotherETB. The switch core can switch up to16Gbps, making it suitable for use as largeHSDPA and E-UL nodes or as a transporthub. As network traffic increases, the switchcore, boosted by the AAL2 switching ca-pacity of the main processor, will play an in-creasingly prominent role.5

E1/T1/J1 interfaces have been integratedinto the CBU subsystem to simplify the

handling of standard, small- and medium-sized RBS configurations.

Mechanics and power

The indoor and outdoor cabinets have beenreorganized to make room for internal powersupplies, the new subsystems, auxiliary sys-tems and improved battery backup. To savespace, reduce power consumption and in-crease availability, the cooling system nowemploys a central fan instead of dedicatedsubrack fans. New fan-control algorithmsmake use of hot-spot measurements, takinginto account unique board characteristics andsound levels. As the subsystems evolve overtime, the fan control subsystem will auto-matically load board characteristics on to newboards and adjust the fan speed accordingly.

The new cooling concept continues to em-ploy the chimney principle of past RBS de-signs (Figure 4). Because no ventilationspace is required, cabinets can be placedside-by-side, back-to-back or back-to-wall.

To further improve availability and re-duce scheduler maintenance, the outdoorcabinet is cooled by means of a heat ex-changer (standard).

The new integrated power option sup-ports +24VDC, -48VDC and mains-supplyAC. This option eliminates the need for anextra power cabinet.

RBS 3000 R3 macro basestationsThe RBS R3 macro base stations comple-ment Ericsson’s existing product portfolio.Considerable effort has been made to ensurecompatibility between the releases. At thesame time, new capabilities have beenadded, including dual-band, larger capaci-ty per cabinet, and improved power effi-ciency. To start with, Ericsson is releasingfive new macro base station cabinets: threeindoor and two outdoor versions.

Indoor cabinets

The three indoor cabinets share similar char-acteristics, such as integrated power supplyand transmission hub functionality. Inessence, they differ only in terms of target-ed maximum configuration.

The RBS 3206E can house nine radiounits for large, dual-band configurationswith very high output power capability.

The RBS 3206F can house six radio unitsand is suitable for high- to very-high-capacity configurations, including dual-band with high-output capability.

Figure 4Cooling principle.


The RBS 3206M targets medium- tohigh-capacity configurations. The standardconfiguration calls for three radio units.

Outdoor cabinets

The RBS 3106 outdoor cabinet has the samefootprint as the GSM 2106 and WCDMARBS 3101. It can be configured in the sameway as the RBS 3206E – that is, with up tonine radio units – and includes an inte-grated power and battery backup system.

Apart from supporting much larger con-figurations than its predecessor (RBS3101), the RBS 3106 now also features a heat exchanger cooling system (standard).

The narrow depth and low height of theslim-sized RBS 3107 gives operatorsgreater flexibility in terms of site acquisi-tion.

RBS R3 key technologies

Power efficiency

Power efficiency has an environmental im-pact and affects operating costs. Lowerpower consumption can reduce costs for en-ergy and reduces the demand charge (con-tract ampere). Ericsson’s life-cycle assess-ments show that reducing RBS power con-sumption goes a long way toward reducingthe total environmental impact of telecom-munications services.6 The assessments con-clude that an energy savings of 1kWh isequivalent to keeping 0.6kg C02 from en-tering the atmosphere.

RBS power efficiency is affected by everypart of the node (baseband, control parts,power supply units, and internal and exter-nal cooling due to heat dissipation) but thedominating factor is power amplifier effi-ciency.

New linearization technology

The main driving factors for introducing a newlinearization technology are the potential • to drastically increase power amplifier

(PA) efficiency; and• to increase RBS capacity in a given foot-

print (increased density).Adaptive baseband digital predistortion(DPD) is a mature technology that has movedfrom research labs into deployed products.When combined with advanced peak powerreduction algorithms, DPD significantly improves efficiency compared to the feed-forward PAs used in earlier releases.

For WCDMA four-carrier operation,power efficiency of the transmitter chain(that is, transceiver and power amplifier) canbe improved from typically less than 10%to around 15%. DPD technology enablesthe active radio parts of the RBS to be inte-grated into a complete radio unit (RU) withdigital baseband input signals.

Adaptive DPD

The baseband signal is predistorted beforemodulation, up-conversion, and amplifica-tion in the power amplifier. Figure 5 showsthe relationship between the PA input sig-nal and output power. The PA curve beforelinearization is nonlinear until it reaches sat-uration. With DPD, the PA curve is forcedto have a linear response over a specific op-erating range. Figure 6 shows a block dia-gram of the complete DPD system.

Before entering the DAC, samples of thebaseband input signal are multiplied bycomplex coefficients drawn from the look-up table (LUT). The LUT coefficients,which implement the predistortion func-tion, are updated according to changes inPA behavior relative to changes in traffic,the environment, and aging effects.


Figure 5

Digital predistortion (DPD) principle.

Figure 6Block diagram of adaptive baseband digi-

tal predistortion (DPD).



Ordinary memoryless DPD algorithmsare not well suited to cope with the PAmemory effects created by rapid dynamicchanges in average power level. To mitigatethese effects while still fulfilling the moststringent 3GPP linearity requirements,Ericsson has developed advanced DPD al-gorithms with fast adaptation. To achieveoptimum efficiency the DPD is combinedwith peak power reduction algorithms thatreduce the signal peak to average value with-out sacrificing error vector magnitude(EVM) properties. The hardware is com-posed of DACs, ADCs and LD-MOS powertransistors that linearize four WCDMA car-riers over a 20MHz operating bandwidth.One can easily adapt the architecture topower amplifiers with different outputpower levels, amplifier technologies, andnew RF power transistor technologies.

The following definitions are necessary forcomparing efficiency values (Figure 7):• power amplifier efficiency includes the

driver and final stages as well as losses inthe PA output network;

• radio unit (RU) efficiency includes theDC/DC converter, the TRX unit, and thePA as defined above.

Measurements show a significant efficiencyimprovement with DPD compared to theRBS R2 with analog feed-forward ampli-fiers. Figure 8 shows the typical measuredRU efficiency versus the Pout curve. Themeasurements, taken at room temperature,measured 30W RBS power using 3GPP testmodel 1 (TM1) signals.

The efficiency at maximum power(46dBm/40W) is 15%. The correspondingDC power consumption is 270W for a com-plete radio unit. By comparison, an RBS R2

Figure 7Efficiency definitions.

Figure 8Typical RU efficiency curve.


feed-forward MCPA with TRX typicallyconsumes 400W at this power level.

Figure 9 shows the measured performanceor adjacent channel power leakage ratio(ACLR) of RU21 at Pmax (40W). The mea-surement was taken using two WCDMAcarriers in 15MHz bandwidth (2162.4MHzcenter frequency), the carriers were modu-lated with a 3GPP TM1 signal. The ACLRand spurious emission responses are wellwithin the stipulated requirements.

Further efficiency enhancements

Proceeding from the proven DPD design,the RBS R3 also supports other efficiency-enhancing technologies and new powertransistor technologies. Doherty technolo-gy, for example, increases the average effi-ciency of a power amplifier with little in-crease in complexity.

In a Doherty amplifier, two amplifiers ofequal capacity can be combined throughquarter wavelength lines. Each amplifier isdesigned to give maximum power at a loadof 50 ohms.

The main PA is biased in Class AB, whilethe peak PA can be biased in Class AB orClass C (Figure 10). When the signal am-plitude is half, or less than half, of the peakamplitude only the main PA remains active;the peak PA is switched off. Each PA con-tributes to the output power when the sig-nal exceeds half the peak amplitude. In re-ality, the main PA load is modulated withchanges in output power.

Figure 11 shows the efficiency curve forthis coupling, using two PAs. Peak effi-ciency is set at 6dB back-off. Other divisionratios may be used to shift the curve to theleft or right to match the actual signal peak-to-average ratio.

Because the Doherty architecture is in-herently non-linear, a good linearizationtechnology is required to fully exploit theefficiency enhancement properties. Theproduct verification measurements haveproven that a Doherty PA combined withadvanced DPD algorithms can meet thestringent 3GPP linearity requirements andstill significantly improve efficiency.

Figure 12 shows that the introduction ofDoherty power amplifiers increases RU ef-ficiency to around 20% at Pmax. The Do-herty effect is very evident when comparedwith the pure DPD PA curve (Figure 8). Alarge improvement in efficiency occurs ataround 6dB below Pmax.

HSDPA and R99 traffic with optimal

capacity

The inclusion of HSDPA in 3GPP Release 5 represents a major improvement inWCDMA capacity, latency and peak rate.Thanks to higher-order modulation, fast re-transmissions and fast link adaptation, thedownlink can attain a maximum bit rate of14.4Mbps with average cell throughput of upto 5Mbps. HSDPA increases the capacity ofthe air interface two- to three-fold, yieldinga much-improved end-user experience. It also


Figure 9Measured adjacent channel leakage

power ratio (ACLR) in the RU21.

Figure 10

Doherty PA principle.



has the potential to improve cost-effective-ness in the radio access network.

An important aspect of HSDPA is that itcan either be enabled on the same carrier asR99 traffic, to make optimum use of carri-er resources, or on a separate carrier, to pro-vide dedicated capacity for mobile broad-band.

The different deployment scenarios putdifferent functional and performance re-quirements on the RBS. In terms of radiodesign (scheduler, power resource alloca-tion, transmitter linearity) it is straightfor-ward to deploy HSDPA and R99 on sepa-rate carriers, which results in a large, dedi-cated resource for mobile broadband. How-ever, in the context of radio resources, de-ploying HSDPA and R99 traffic on the samecarrier is an attractive option, because thecarrier can be used as a common resourcepool for high-speed, best-effort data anddedicated voice and data traffic.

Given that R99 traffic consists of fixedcommon channels and power-controlleddedicated channels, the carrier must be di-mensioned with an output power marginthat can handle varying instantaneouspower demands. Resources go unused when-ever the output power falls below the nom-inal output power. HSDPA can employ un-used output power without any negative im-pact on R99 traffic. The HSDPA traffic issimply allocated whatever power is availableafter the R99 demand has been met.HSDPA power allocation is updated dy-namically every 2ms (Figure 13).

Efficient output power handling for HSDPA

Efficient management of output power re-sources for a common HSDPA and R99 car-rier is dependent on a variety of parameters,including• dynamic output power allocation;

• TX chain linearity; and• fast congestion control. As mentioned above, an effective imple-mentation dynamically allocates all excessoutput power to HSDPA (Figure 13). Thesystem updates the HSDPA output powerevery 2ms. By contrast, it updates R99 radiolink power in increments of 1dB every0.67ms. R99 power-controlled traffic canthus request an increase of up to 3dB in out-put power before the HSDPA power settinghas been updated a second time. In a sce-nario with 8W average power-controlledR99 traffic, this 3dB increase is the equiv-alent of 8W. If all temporarily available out-put power were to be allocated to HSDPA,then fluctuations in R99 power might over-drive the TX chain, resulting in modulationerrors, spectrum emissions or even forcedMCPA shutdown. A standard solution toprevent this from occurring is to restrict thepower allocation to HSDPA by means of anHSDPA power margin. However, HSDPApower margins lower throughput for bothHSDPA and R99 traffic. Ericsson’s solutionis to employ an RBS downlink fast conges-tion control mechanism which ensures thatthe TX chain is never overdriven by fluctu-ating R99 traffic. This way, full power ca-pability is available for HSDPA and R99traffic.

TX chain linearity

In oversimplified terms, an amplifier designis based on peak and average power re-quirements. The maximum average poweris determined by the cooling design, and thepeak power is determined by linearizationperformance and allowable spectrum emis-sions. The peak-to-average ratio (PAR) hasa direct impact on power efficiency. LowPAR yields a more efficient amplifier. Thisis why peak clipping functions are used to

Figure 11

Ideal Doherty PA efficiency curve.

Figure 12

Efficiency vs. Pout curve obtained from RUwith prototype Doherty PA.


hold power peaks down while maintainingadequate modulation accuracy. The intro-duction of HSDPA requires higher-ordermodulation, which• infers higher PAR; and• requires more accurate modulation.The clipping algorithm sees higher peaksbut must have less impact on the modula-tion waveform. The clipping must thus bedesigned with 16QAM modulation inmind. Otherwise, to meet requirements formodulation accuracy, one must back the sig-nal off. Also, because increased signal peaksinfluence the spectrum emission of thepower amplifier, its linearity must be pre-pared for HSDPA.

In comparison with R99, the higher-order modulation (16QAM) of HSDPA putsmore stringent requirements on transmitterchain linearity and the performance of clip-ping algorithms. A strictly 3GPP R99-compliant transmitter chain would requirepower back-off of approximately 1.6dB(30%) to meet the tougher requirementsimposed by HSDPA. In the 20W powerclass, this would be equivalent to an unde-sirable 6W drop in nominal output poweror a 30% drop in cell capacity. To avoid thispower back-off when introducing HSDPA,Ericsson designed its RBS products to havelarger dynamic range and better linearityand modulation accuracy than stipulated by3GPP R99.

Fast congestion control

The two central, shared resources in thedownlink are the code tree and outputpower. The admission control functionserves to ensure that admitted users enjoy ahigh likelihood of obtaining the outputpower and codes needed for their services.The congestion control function takes ac-tions to lower output power when the car-rier power level exceeds a given threshold.Poor congestion control (weak, slow re-sponse) calls for severe actions. Efficient con-gestion control (robust, fast response) callsfor less severe actions, in which case themean power can be maintained at a higherlevel. This also means that more users canbe admitted into the cell. There is thus a di-rect relationship between the congestioncontrol function and cell capacity.


Figure 13Fast congestion control.

Figure 14

HSDPA traffic utilizes available output power.



Fast congestion control (FCC) is an RBSfunction that complements RNC congestioncontrol. The function supervises the outputpower that users (all users) demand at thesame time, using the same time scale as thefast power control function. If the total de-mand for output power exceeds nominal out-put power, the total carrier power is heldsteady at nominal output power until theRNC congestion control function has takenenough corrective actions, for example, byswitching down the data rates. The reactiontime of the FCC function matches that of R99power control (0.67ms), which is to say it isfast enough to fully prevent saturation of theTX chain or overdriving of the power am-plifier without the need for power margins.Therefore, cell behavior remains robust atmaximum load without running the risk ofdropped cells or modulation inaccuracy. Fur-thermore, the RNC congestion and admis-sion thresholds can be set to higher levels, in-creasing cell capacity without compromisingoverall quality of service (Figure 12).

Ericsson estimates that FCC yields up to25% greater capacity for R99 traffic. ForHSDPA, the gain is nearly 50% greaterthroughput during the “busy hour.”

ConclusionEricsson’s strategy for 3G network develop-ment is to release products and features inaccordance with customer needs at differentphases of network rollout.

The first indoor and outdoor macro basestations were released in early 2001. Theseproducts were based on the first commercialRBS design release (RBS R1).

The second RBS design release (RBS R2) improved on the architectureand subsystems in RBS R1 and introducedsoftware support for additional configu-rations. Ericsson also broadened its RBSp o r t f o l i o .

The new third release of EricssonWCDMA radio base stations employs tech-nology advances to improve the architec-ture of the RBS node to help operators meetchanging market requirements and targetnew business opportunities. The new design enables a doubling of the node capacity, increased coverage, simplifiedmaintenance, and dramatically reducespower consumption. Taken as a whole theseenhancements help operators to keep theirradio access network-related CAPEX andOPEX low.

1. Reinius, J.: Cello – An ATM transport and

control platform. Ericsson Review, Vol.

76(1999):2, pp.48-55

2. Bjärhov, M. and Friberg, C.: GSM network

solutions for new-growth markets. Erics-

son Review, Vol. 78(2001):1: pp. 6-15

3. Zune, P.: Family of RBS 3000 products for

WCDMA systems. Ericsson Review, Vol.

77(2000):3, pp. 170-177

4. Zhang, Z., Heiser, F., Lerzer, J. and

Leuschner, H.: Advanced basedband tech-

nology in third-generation radio base sta-

tions. Ericsson Review, Vol. 80(2003):1,

pp. 32-41

5. Karlander, B., Nádas, S., Rácz, S. and

Reinius, J.: AAL2 switching in the WCDMA

radio access network. Ericsson Review,

Vol. 79(2002):3, pp. 114-123

6. Edler, T. and Lundberg, S.: Energy effi-

ciency enhancements in radio access net-

works. Ericsson Review, Vol. 81(2004):1,

pp. 42-51

7. Berglund, B., Nygren, T. and Sahlman, K-

G: RF multicarrier amplifier for third-gener-

ation systems. Ericsson Review, Vol.

78(2001):4, pp. 184-189

REFERENCES

2G Second-generation mobile system

3G Third-generation mobile system

3GPP Third Generation Partnership

Project

AAL2 ATM adaptation layer 2

ACLR Adjacent channel leakage power

ratio

ADC Analog-to-digital converter

ASC Antenna system controller

ASIC Application-specific integrated

circuit

ATM Asynchronous transfer mode

CAPEX Capital expediture

CBU Control base unit

CDMA Code-division multiple access

CE Channel element

CPP Connectivity packet platform

DAC Digital-to-analog converter

DL Downlink

DPD Digital predistortion

DSP Digital signal processor

ETB Exchange terminal board

E-UL Enhanced uplink

EVM Error vector magnitude

FCC Fast congestion control

FPGA Field-programmable gate array

FU Filter unit

GPB General-purpose board

GSM Global system for mobile

communication

HSDPA High-speed downlink packet access

ISP In-service performance

LCA Life-cycle assessment

LD-MOS Lateral double-diffused metal-oxide

semiconductor

LUT Look-up table

MCPA Multicarrier power amplifier

O&M Operation and maintenance

OPEX Operating expense

PA Power amplifier

PAR Peak-to-average ratio

PDH Plesiochronous digital hierarchy

PIU Plug-in unit

QAM Quadrature amplitude modulation

R&D Research and development

RAN Radio access network

RAXB Receiver and random access board

RBS Radio base station

RET Remote electrical antenna tilt

RF Radio frequency

RNC Radio network controller

RU Radio unit

SDH Synchronous digital hierarchy

TCO Total cost of ownership

TM1 Test model 1

TMA Tower-mounted amplifier

TX Transmitter

TXB Transmitter board

UL Uplink

WCDMA Wideband CDMA

BOX A, TERMS AND ABBREVIATIONS


third design release of ericsson’s wcdma macro radio … phases background ericsson’s strategy...

Documents