Capacity solutions for GSM radio access net-works continue to be in focus. In the Amer-icas, non-GSM technologies, such as TDMAand AMPS, occupy a substantial amount ofspectrum. Likewise, competition makes itdifficult for operators to acquire new spec-trum. Therefore, operators who migrate toGSM are faced with the challenge of pro-viding enough capacity for new serviceswhile maintaining capacity in legacy sys-tems. The downturn in the world economyhas also put constraints on operators, forc-ing them to maximize benefits from everyinvestment. Since high-capacity solutionsare about building high-capacity networksin the most economical way, GSM radio net-work capacity solutions are perhaps moreimportant today than ever before.

Radio network capacity solutions can bedivided into three solution categories: • cell capacity solutions—these solutions

consist of methods and features that per-mit more transceivers per cell;

• network capacity solutions—these solu-tions focus on adding different kinds of

cells and on making the most of cell ca-pacity by distributing traffic as efficient-ly as possible; and

• channel capacity solutions—these solu-tions center on ways of using the availablethroughput of the channels in the air in amore efficient manner, for example half-rate voice channels and GPRS.

Cell capacityThe one factor that has the greatest influ-ence on cell capacity is frequency reuse. Cellcapacity is thus determined by differentmethods and functions to enhance frequen-cy reuse. Two common methods are • multiple reuse pattern (MRP); and • fractional load planning (FLP). The multiple reuse pattern, which is basedon baseband frequency hopping, yields thebest results for networks composed mainlyof filter combiners. The primary transceiv-er carries the broadcast control channel(BCCH) and must therefore have a relative-ly loose reuse pattern (explanation: a hand-set must listen to the information broadcaston the BCCH before it can make calls in acell). But thanks to the frequency hoppinggain, all remaining transceivers in the net-work can have a successively tighter reusepattern. Compared to a non-hopping net-work, the MRP solution can more than dou-ble cell capacity. The drawbacks of MRP arethat it requires • considerable spectrum (greater than

7 MHz); and • at least three transceivers per cell for good

performance.Fractional load planning is based on syn-thesized frequency hopping, which requiresthe use of hybrid combiners. In FLP, the gainfrom frequency hopping is not dependenton the number of transceivers in a cell, sinceeach transceiver can hop on every frequencyallocated to the cell. Notwithstanding, dueto the characteristics of synthesized fre-quency hopping, the BCCH transceiver can-not hop frequencies. Ordinarily, to guaran-tee adequate voice quality for a non-hoppingtraffic channel (TCH), a frequency reuse ofapproximately 15-18 is needed. But byusing the BCCH in an overlaid subcell fea-ture, it is possible to plan BCCH frequencyreuse as if the BCCH transceiver could hopfrequencies, making a frequency reuse of ap-proximately 11-12 feasible. The most com-mon FLP methods in use are 1/3 and 1/1,and FLP can be implemented in frequencybands as narrow as 3 MHz.

84 Ericsson Review No. 2, 2002

Ericsson’s GSM RAN capacity solutionsPeter Blom

High-capacity solutions are about building high-capacity networks in themost economical way, and therefore, GSM radio network capacity solu-tions are becoming increasingly important. Radio network capacity solu-tions can be divided into three categories: cell capacity, network capacity,and channel capacity.

The author discusses various solutions for improving radio networkcapacity in each of these areas. He also describes different implementa-tions and recommends three general steps for introducing solutions in thenetwork.

AMPS Advanced mobile phone systemAMR Adaptive multirateBCCH Broadcast control channelDTX Discontinuous transmissionEDGE Enhanced data rates for global

evolutionEFR Enhanced full-rateFAS Frequency allocation supportFLP Fractional load planningFR Full-rateGPRS General packet radio serviceGPS Global positioning systemGSM Global system for mobile

communication

HR Half-rateIRC Interference rejection combiningMAIO Mobile allocation index offsetMRP Multiple reuse patternPCCCH Packet-data common control

channelQoS Quality of serviceRBS Radio base stationTCH Traffic channelTDMA Time-division multiple accessTET Traffic estimation toolWCDMA Wideband code-division multiple

access

BOX A, TERMS AND ABBREVIATIONS

85

A third method, known as non-uniformfrequency planning, can be a mix of MRPand FLP, or a totally free allocation of fre-quencies. Since this method is more com-plex, cell-planning and measurement toolsare recommended, such as TEMS Cell Plan-ner and frequency allocation support(FAS).

Ericsson’s GSM system provides a host offeatures that minimize frequency reuse andmaximize cell capacity. The most basic fea-ture—frequency hopping—is unique toGSM and is the basis for all tight frequen-cy reuse solutions. Since a cell can accom-modate up to 16 different hopping or non-hopping frequency groups (channel groups),there is considerable inherent flexibility foradapting the network to different serviceswith different requirements for radio qual-ity. Radio base station (RBS) site synchro-nization and mobile allocation index offset(MAIO) management are provided to max-imize the benefits of FLP. MAIO is the pa-rameter that determines when a frequencyis used in a cell that employs FLP. Withproper MAIO planning it is possible to min-imize or even eliminate interference be-tween synchronized cells.

Similarly, quality-based dynamic powercontrol and discontinuous transmission(DTX) are employed to minimize radio in-terference. The quality-based power controlfeature performs very well in networks thatemploy tight frequency reuse—it providesa substantial decrease in output power, es-pecially compared to non-quality-based al-gorithms. Should quality-related problemspersist, an intra-cell handover function findsa new channel in the cell on which the callcan continue.

In addition, the dynamic overlaid/under-laid subcells feature divides the cell into twosubcells with traffic management function-ality between them. It also makes it possi-ble to restrict the coverage of the overlaidsubcell, thereby facilitating even tighterreuse in it. And finally, there is the adaptivemultirate (AMR) voice codec for GSM full-rate channels. The feature uses several voicecodecs and associated error protection (chan-nel coding) to adapt to the quality of theradio environment. Compared to the full-rate (FR) and extended full-rate (EFR) voicecodecs, this feature greatly improves ro-bustness. In fact, compared to EFR, whenAMR full-rate is used to its fullest extent—to tighten frequency reuse and to add moretransceivers—the traffic capacity in the cellcan be doubled.

Network capacityIn addition to improving cell capacity, op-erators can introduce micro cells, since siteacquisition for micro cells is usually easierand less expensive than when adding regu-lar cells. To facilitate the deployment ofmicro cells, Ericsson provides the traffic es-timation tool (TET), which enables opera-tors to identify the optimal site location.And since it is possible to measure theamount of traffic a certain site location willcarry, he can then calculate time to paybackbefore the operator installs the base station.

A multiband network option is open tooperators who have frequencies in a secondfrequency band. The amount of capacity thatcan be derived from implementing cells ina second frequency band depends on theamount of spectrum in the alternative fre-quency band. Even so, since the cell capac-ity solutions described above can be appliedin any frequency band, the extra capacity isnever negligible. To derive the optimumcost-benefit ratio, Ericsson recommendsthat the operator reuse all existing sites forRBSs that belong to the new frequencyband.

Traffic management is an important issuein a network composed of cells of different

Ericsson Review No. 2, 2002

Voice quality

Channel quality (C/I)

Mode 1

Voice codecmode changes

AMR performance

Mode 2

Mode 3

Figure 1The principles of AMR. Voice quality is maximized by adapting—or by switching betweenseveral voice codecs—to the quality of the radio channel.

sizes and frequency bands. Intelligent traf-fic distribution algorithms let cells cooper-ate and help one another to enhance networkcapacity to a degree that exceeds the sum ofall individual cells.

With multilayered hierarchical cell struc-tures—the most important traffic-handlingfunction in Ericsson’s GSM system—cellscan be divided in up to eight layers, and traf-fic can be prioritized and distributed be-

tween these layers. There are also numerousadd-on functions, such as • cell load sharing, which distributes traffic

within layers; • assignment to another cell, which redirects

traffic to other cells when congestion oc-curs during call setup; and

• handling of fast-moving mobiles, whichmoves calls to higher layers when thereare too many handovers within a given in-terval. This function reduces the numberof handovers, thereby increasing voicequality and reducing processor load.

Channel capacityIn the context of circuit-switched traffic,channel capacity is about half-rate voicechannels and the way they are managed.Since the half-rate technique reduces thequality of voice, it has not been widely de-ployed. However, operators are now begin-ning to use this technique more and more,since it can be allocated on a dynamic basis(dynamic HR allocation) during trafficpeaks.

In the context of data communications,GPRS is a channel capacity solution. Itmakes optimum use of channels and maxi-mizes capacity by allowing several users toshare the same channels. Ericsson’s GPRSsolution provides dedicated as well as on-demand packet-data channels. The solutionalso supports dedicated packet-data com-mon control channels (PCCCH). During2002, numerous improvements will bemade available, including EDGE and

86 Ericsson Review No. 2, 2002

Figure 2Network capacity: Efficient traffic-management features, such as Ericsson’s multilayeredhierarchical cell structure (HCS), should be employed when different kinds of cells and fre-quency bands are used.

Figure 3Dynamic half-rate allocation: To mitigatethe impact on voice quality, half-ratetechniques should only be used duringtraffic peaks.

Ericsson Review No. 2, 2002 87

quality-of-service-based (QoS) scheduling.EDGE technology, which can more thantriple throughput per channel, enhancespacket data capacity and facilitates a multi-tude of new services that require extra band-width. The QoS scheduling functions allowoperators to differentiate their service offer-ing to distinct user segments.

Capacity gainsFigure 4 compares the most common op-tions for improving network capacity. Ascan be seen, most of the methods have thepotential to double network capacity, andour recommendation is that the operatorshould implement or apply as many of theseas possible. In particular, two solutions re-ally stand out: AMR full-rate, and microcells. Note: To derive the greatest gains fromAMR full-rate, the penetration of AMRhandsets must be high.

Implementation aspectsApart from gains in capacity, the two mainparameters that an operator should consid-er when building a network are monetarycost and time—the actual cost of each solu-tion is market-dependent, since the costs as-sociated with cell sites (site acquisition, sitepreparation, rental costs) and transmissionvary from market to market.

In every market it costs time and moneyto build sites. Accordingly, the greater thenumber of sites required, the higher the cost.When viewed in this light, we can concludethat tight frequency reuse offers the mostexpedient and cost-effective solution to im-proving capacity, since in many cases the op-erator needs only add transceivers to exist-ing cabinets.

If the operator wants to maximize his useof existing sites then the second-best optionis to deploy transceivers on other availablefrequency bands. In this case, the operatorneeds only add cabinets at sites where extracapacity is wanted.

A third option is to introduce microcells—thanks to the small size of micro basestations, it is easier and less expensive to ac-quire sites for them.

A final option is to build more sites.As mentioned above, the operator might

also make use of half-rate channels, but sincethis option decreases voice quality, it shouldbe allocated dynamically (dynamic HR al-location), and then mostly as a last-resortoption when network expansion through

other means cannot keep pace with growthin traffic. The half-rate technique can be de-ployed very quickly by activating varioussoftware features.

Recommended way ofincreasing capacityTaking into account the potential for ca-pacity and the implementation aspects for

Relative capacity gain (Erlang)

0

1

2

3

4

5

6

7

8

Ref.network

Tightmacrocells

Tighterfreqencyreuse

Multi-band

Microcells

Half-rate

Figure 4Reference network: Average networkwithout frequency hopping. Tight macro cells: Double the number ofmacro cells. Tighter frequency reuse: (a) MRP networkbased on EFR; (b) FLP 1/1 network basedon EFR; (c) FLP 1/1 network based onAMR.Multiband: (a) 50% penetration of capableterminals; and (b) 100% penetration ofcapable terminals.Micro cells: One micro cell every 200 m.(a) 2 TRX per micro cell; (b) 4 TRX permicro cell.Half-rate: (a) 25% penetration of capableterminals; (b) 100% penetration of capa-ble terminals.

Cost for doubling radio network capacity [%]

0

20

40

60

80

100

Tightmacrocells

Tighterfreqencyreuse

Multi-band

Microcells

Figure 5Cost comparison: The additional cost ofdoubling radio network capacity usingvarious solutions relative to the originalinvestment.

each capacity-boosting solution, we can rec-ommend a general order for introducing thedifferent solutions in the network.

Step 1Operators should activate frequency hoppingand implement one of the tighter frequencyreuse methods. Which method should beused depends on factors such as RBS hard-ware, amount of spectrum, and cell plan. Ad-ditional reuse-enhancing features might alsobe necessary depending on the reuse methodand the degree of frequency use.

Operators should activate AMR. Al-though the initial gain in capacity is small,a significant improvement in quality can beachieved for users with AMR-capable ter-minals. As the penetration of AMR hand-sets increases it will allow more transceiversto be deployed in the network, gradually en-hancing traffic capacity.

Step 2Operators who have access to spectrum inother frequency bands should start deploy-ing equipment to use those frequencies.Costs can be kept to a minimum if opera-tors reuse existing sites.

Operators can add micro cells or indoorcells at traffic hot-spots, which include pop-ular squares, conference centers, shoppingmalls and airports. The use of micro cells tocover hot-spots will offload the macro cells,and help operators to avoid the cost of hav-ing to split cells.

Operators can also use dynamic half-rateallocation to avoid congestion at peak traf-fic. This measure reduces the pressure onoperators to build out the network, allow-ing them to build it out at a manageablepace.

Step 3As traffic increases, the number of microcells and indoor cells will also continue togrow. At some point, the micro cell layerwill be almost contiguous. By adding morecapacity to the micro cells and indoor cells,operators can achieve an extreme boost incapacity.

Future GSM RANcapacity solutionsGSM has been in commercial operation forsome 10 years now, but there is still ampleroom for enhancing capacity. One area thatshows great potential is the use of advancedinterference-suppression algorithms. In thebase station, we will see this type of algo-rithm introduced in the form of interfer-ence rejection combining (IRC). The interference-suppression performance ofthis feature far outperforms present-day re-ceiver algorithms. Simulations show thatIRC can give link gains of up to 5 dB innon-synchronized networks, and up to dou-ble that in synchronized networks. This canbe translated into better voice quality anddata throughput on the uplink. However,due to a lack of corresponding functionali-ty in present-day handsets, the gain in ca-pacity will be limited. To remedy this sit-uation Ericsson is also researching interfer-ence suppression algorithms for terminals.Limited processing capacity, small size, andthe importance of design (for example, onlyone antenna) impose harsh restrictions onalgorithms. However, Ericsson has madesome technological breakthroughs and ex-pects to introduce powerful interference-suppression algorithms in handsets in thenext few years.

88 Ericsson Review No. 2, 2002

Figure 6Micro base station RBS 2302.

Ericsson Review No. 2, 2002 89

Apart from boosting IRC performance,synchronized radio networks also enhancecapacity. Synchronization is achieved bysynchronizing all RBS sites to the GPS sys-tem. When all cells are synchronized to thesame reference clock, interference planningwill no longer be limited to cells located atthe same site. Instead, operators will be ableto determine which cells truly interfere witheach other, no matter where they are locat-ed. Here, too, simulations show gains of upto 20% compared to a site-synchronizednetwork.

AMR technology will also enhance thechannel capacity solution for half-rate chan-nels: AMR half-rate consists of a subset ofthe AMR voice codecs defined for full-ratechannels. As with AMR full-rate, it adapts,by switching between voice codecs, to thequality of the radio environment. AMR half-rate improves voice quality compared to thepresent voice codec for half-rate channels,making the half-rate option a more attrac-tive solution for increasing capacity.

Looking further ahead, better techniquesfor distributing traffic between GSM andWCDMA will be introduced until we cansee a seamless GSM-WCDMA network,which will allow operators to use both tech-nologies to their fullest, to maximize theend-user experience. It will also facilitate theintroduction of new functionality in GSM.The introduction of adaptive antenna tech-nology will yield even greater gains. Ericsson has tested adaptive antennas ex-tensively to verify their performance. Thesetests show that adaptive antennas have thepotential to more than double network ca-pacity. In fact, it will be nearly doubled ifadaptive antennas are installed at only 20-30% of the sites in a given area.

ConclusionRadio network capacity solutions can be divided into three categories: cell ca-pacity, network capacity, and channel capacity.

The one factor that has the greatest in-fluence on cell capacity is frequency reuse.That is, cell capacity is determined by dif-ferent frequency reuse methods and func-tions to enhance it. Two common methodsare multiple reuse pattern and fractionalload planning. A third frequency reusemethod is non-uniform frequency plan-ning.

To improve network capacity, operatorscan introduce micro cells, multiband oper-ation (if additional spectrum is available),traffic management, and multilayered hier-archical cell structures.

Dynamic half-rate allocation technologyenables operators to increase channel capac-ity for circuit-switched traffic during trafficpeaks. Likewise, GPRS is a channel capaci-ty solution for data communication—itmakes optimum use of channels and maxi-mizes capacity by allowing several users toshare the same channels.

A recommended three-step approach toincreasing capacity in the radio access net-work is as follows:1.Activate frequency-hopping, employ

tighter frequency reuse, and activateAMR.

2.If spectrum is available in a second band,deploy equipment in those frequencies.Add micro or indoor cells at hot-spots,and use dynamic half-rate allocation to re-duce congestion during peak traffic.

3.Add more micro and indoor cells and in-crease the capacity in them.