Wireless in Local Loop — Some Fundamentals

Ashok Jhunjhunwala, Devendra Jalihal, K.Giridhar
Telecommunications and Networks (TeNeT) Group
Dept. of Electrical Engineering, IIT Madras
Chennai 600 036, INDIA



Abstract

The enhancements in Internet technology and mobile access technology over the last decade can be leveraged effectively to build Wireless in Local Loop (WiLL) systems, which can enable rapid expansion of telecom and Internet access in developing countries. However, the design of a WiLL system requires one to understand some fundamentals concerning the Access Network and its connectivity to backbone network as well as the traffic requirement for a voice and Internet connection. The requirements of WiLL, in contrast to that of a Mobile cellular system need to be clearly understood. Equally important is the concern for capacity and spectral efficiency, especially as higher bit-rate Internet systems becomes a must for developing countries to get a fair share of the economic advantages that telecom technologies provide. This paper looks at these fundamental issues in context of GSM, IS-95 and DECT technologies. The paper further takes a brief look at some recent technological developments, which are likely to impact the Wireless in Local Loop systems.

The paper concludes with a discussion on the emerging third generation (3G) wireless standards, and the new technologies which are being introduced into the network, and what will be their impact on Internet and multimedia bit-rates and services.

1 Introduction

Till around mid 80s' in India, a local loop or an Access Network (AN) used to consist of a pair of copper wires connecting the subscribers at home or office to the nearest exchange. The local loop length in urban areas would be typically as large as 6 to 8 kms and the copper gauge used was 0.5 mm to 0.6 mm. The loop was designed to carry 4 kHz voice and was difficult to maintain, with almost 85% of all faults found in the local loop. Above all, it was expensive, difficult, and time-consuming to deploy. With the rising cost of copper and cost of digging increasing every year, if one had continued with such an approach, the per line local loop cost itself would have been Rs.40,000 to Rs.50,000 and would have amounted to over 85% of the total cost of putting a telecom network.

Fortunately, an uncelebrated but major technological innovation changed the Access Network from mid-eighties onward. As shown in Fig. 1, the Access Network now consists of a fibre from an exchange to a RLU/RSU and a typically 3-4 km copper loop from the RLU/RSU to the subscriber premises. The signals carried on fibre is time multiplexed digital voice and signalling. A RLU typically serves 1000 to 4000 subscribers, and the signal from RLU to exchange consists of 4 to 16 E1 [1]. The copper used now is only 0.4 mm and the costs are down considerably.

The rising cost of copper however continues to make even this solution expensive. Today the per line 3-4 km copper cost (including laying charges) would be Rs.13,000 to Rs.16,000, the shared fibre cost would typically be Rs.1000 per line, and RLU cost would be Rs.3000 to Rs.4000. The Rs.20,000 plus cost of the Access Network today again amounts to almost 2/3 of the total per line cost.

The signalling protocol on the Access Network (in the signalling slots on E1 link between RLUs and exchange) remained proprietary for about a decade, forcing an operator to purchase Access RLU and exchange from the same manufacturer. In mid 90s', however, access signalling protocols were finally standardised internationally in the form of V5.1 and V5.2 protocols [2].

This separated AN from an exchange and AN could now be independently deployed. The Access Network could now use new innovative technologies like fibre, wireless, DSL on copper, hybrid fibre-coaxial cable or even power-lines. As access dominated the cost, was most fault-prone, and was most time consuming to deploy, availability of new access solutions became the key to expanding the telecom network, specially in developing countries.

This paper focuses on some of the fundamental issues involved in the choice of wireless Access Networks and their interconnection to the PSTN and Internet network from the point of view of developing countries. Obviously, wireless ANs, just like any other access network of today, must connect to an exchange using V5.2 access protocols. The important thin3g is that today a telecom network no longer can focus on providing telephone service, but must integrate Internet services. Section 2 of this paper therefore looks at some of the issues involved in use of Internet on existing telecom networks and the lessons for emerging wireless Access Network. We then proceed to discuss in Section 3, the distinction between wireless in local loop and mobile communication system and the requirements of the two. In Section 4, we look at some important issues governing choice of wireless access technology, the issues that determine capacity and spectral efficiency. In Section 5, we briefly describe some of the key wireless standards that have emerged over the last twenty years and discuss their suitability for use in wireless in local loop. We conclude with a refocus on cost, as reduction of cost is key to expansion of telephone and Internet network in developing countries.

2 The Internet Tangle

Internet has emerged as second only to telephone in connecting people and may tomorrow subsume the telephone network [3]. But today, the Internet access at homes and offices is based on the telephone network. The Internet access today appears to be simple; just take a telephone line, connect a modem and a computer and dial an Internet Service Provider (ISP). The ISP has a bunch of telephone lines and an equal number of modems connecting the users to a Router as shown in Fig.2. This router is connected to other routers on Internet. A dial-up connection to an ISP router gives a user access to everyone and everything on Internet.

This straightforward looking access to Internet, however, has problems. The telephone network is designed to handle 0.1E traffic per subscriber. This is generally adequate for telephony. However, Internet sessions are usually of long duration, very often even exceeding an hour. As a significant percentage of telephone users start using Internet, the load on the telecom network would far exceed 0.1E per subscriber, resulting into severe congestion and eventual collapse. If this has not happened so far, it is only because a small percentage of telephone users have started using Internet.

The second problem is associated with the local call charges associated with using Internet in this manner. The telephone call for Internet costs Rs.26 per hour in Indian cities today, in addition to the charges payable to ISPs. Thirdly, the analog modem to modem link between the subscriber and the ISP is unreliable. One does get 33.6 kbps connectivity sometimes, but connection could go down to 9.6 kbps and even 4.8 kbps at times. Further, the connection often drops. Finally, an ISP with N telephone lines, N modems and a N port router could serve at most N subscribers at time. A connection drop may not get a reconnection during busy hours.

This Internet tangle requires a different approach in order to support future growth. Fortunately, even though an Internet connection is kept on for long hours, a peculiarity of computer to computer communication is that the use of the connection is not continuos but bursty. Packets are transmitted to and from Internet in bursts, with the communication almost silent most of the time. A circuit switched connection on telephone network, however, is unable to take advantage of this and occupies resources throughout the connection, and thereby congesting the network.

An ideal solution to this problem is to have totally shared, packet-switched access. But local loop is usually a separate physical line to each subscriber and packet access on this unshared line gives no advantage, as no one else can use this resource. In such a situation, it is advisable to separate the Internet data at a point nearest to the subscriber, where data from multiple subscribers can be multiplexed. This is shown in Fig.3, where separation of Internet data and voice data takes place at the Access Unit (AU), located typically at a street-corner.

As shown in the figure, both wired and wireless interfaces to the AU are possible. High-speed digital subscriber loop technology (HDSL) and narrow band ISDN equipment can provide high speed, simultaneous reliable voice and Internet access on a single copper pair. In wired access there is strictly no restriction of the bit-rate between the subscriber and the AU.

However, wireless access makes use of an important shared resource, the frequency spectrum. It is this resource which limits the capacity of a wireless system. Therefore medium access strategies which assigns channels to a subscriber only when he/she wishes to transmit a packet would significantly enhance capacity for Internet access. Wireless access networks which can share the frequency spectrum and utilise it during packet burst are obviously very attractive candidates for rapid expansion of Internet access in the future.

3 Wireless in Local Loop Vs Mobile Wireless Access System

There is little doubt that wireless access systems deployed at the turn of century would provide digital access. Wireless connectivity to subscribers today is provided by mobile communication systems as well as wireless in local loop systems. These two appear to be similar and are often confused with each other. However, the requirements for the two systems are significantly different.

3.1 Mobile Telephone System

Mobile Telephone systems is primarily meant to provide telephony for people on the move. The telephone is meant to keep the person connected while he/she is away from home and office. The key here is universal coverage. The mobile telephone must be reachable wherever the subscriber is, in the car, on the street, or in a shopping mall. Other requirements are less severe. A modest voice quality is acceptable as the user may mostly be speaking from a location with high ambient noise. Data communication is not very important, and wherever required, low-bit rate data communication will be acceptable. Fax communication is unlikely to be used. Furthermore, the traffic per subscriber will not be very high, since the user is unlikely to make long calls. One is typically looking at 0.01 to 0.02 Erlang traffic per subscriber. Some air-time charges for such premium service is generally acceptable.

3.2 Wireless in Local Loop System

Wireless in Local Loop (WiLL), on the other hand, is meant to serve subscribers at homes or offices. The telephone provided must be atleast as good as wired phone. Its voice quality must be high -- a subscriber carrying out long conversation must not be irritated with quality; one must be able to use speakerphones, cordless phones and parallel phones. The telephone must support fax and modem communications and should be connectable to a Public Call Office. Ability to provide atleast medium rate Internet access is a must. Further, the traffic supported should be reasonably high — at least as high as 0.1E per subscriber. Besides, ability to support a large number of subscribers in an urban area (large teledensity) with a limited frequency spectrum is required. Finally, for the systems to be of use in developing countries, the cost of providing this wireless access should be less than that required for wired telephone. Air-time charges are totally unacceptable.

Therefore, eventhough the mobile communication systems and wireless in local loop systems may appear to be similar, and sometimes even used interchangingly, the requirements are quite distinct. Let us now take a more detailed look at some of the issues governing the choice of wireless access technology.

4 Capacity and Spectral Efficiency

Having looked at the way to interconnect a WiLL system to the PSTN and the requirements that a WiLL system has to fulfill, let us now take up the most important issue that governs the choice of a WiLL technology. A wireless communication system has to recognize that the frequency spectrum available will always be limited. Obviously, since the telephone as well as a Internet connection is not used continuously, the channels must be assigned to a subscriber on demand. But this is not enough. The key focus has to be efficient use and reuse of the spectrum.

What governs use and reuse of spectrum?

The use and reuse of spectrum is governed by multiple factors including:

    1. channel pay load (bit rate)
    2. signalling overhead
    3. modulation efficiency
    4. cell-radius (range)
    5. choice of multiple access
    6. interference reduction techniques
    7. spatial diversity and space-time processing

We will discuss spatial diversity and space-time processing in section 6. Let us discuss the other factors here.

4.1 Channel Pay Load

It is obvious that higher bit-rate payload will require larger frequency resources as compared to a lower bit-rate payload. Therefore for voice communication on wireless systems, it may be desirable to have efficient voice compression and lower bit-rate voice codecs. The resulting slightly inferior quality is quite acceptable for mobile communications. But for telephones at homes and office, toll quality voice communications at 32 kbps / 64 kbps may often be desirable. Besides PCM and 32 kbps ADPCM give large degree of transparency for other communication services like fax communications. Even then, lower bit-rate voice communications may sometimes be acceptable. However, when one wishes to also use the line for Internet communications, higher bit-rate communication will obviously be desirable. As the frequency resource used per channel is directly proportional to the payload bit-rate, medium to high bit-rate Internet implies higher use of frequency resource.

4.2 Signalling overhead

As signalling is key to setting up, monitoring and tearing down of a call, signalling communications need to be carried out on air between subscriber equipment and the base stations. The signalling channels may be dedicated for each user or may be shared. Usually, more sophisticated the system, more is the signalling requirement. The signalling becomes an overhead that takes away certain frequency resources and plays a role in overall efficiency of spectrum usage.

 

4.3 Modulation Efficiency

The modulation technique employed has a direct bearing on efficient use of spectrum. Highly spectrum efficient techniques have been developed over the years. For example, 16-QAM technique is more spectrally efficient compared to 8-QAM technique, which in itself is more efficient than QPSK and MSK modulation techniques. But more efficient techniques are usually expensive to implement and may sometime require larger power margins. These techniques are used commonly with systems such as high bit-rate point to point microwave links as the number of such systems required are small and each of these systems is shared by a number of subscribers. But for wireless in local loop, one would require the technique to be implemented in each subscriber’s equipment. Therefore cost in an important consideration. Further, often the power margins available is not large. Therefore QPSK, MSK or even BFSK techniques are often used eventhough their spectral efficiency is moderate.

4.4 Cell radius

Cell radius is perhaps the most important factor governing the spectrum utilisation in a wireless system. Let us take a simple example. Let there be N independent channels available for use in a cell of radius r. It is the reuse efficiency discussed later in
section 4.6, which would determine the reuse of channels in neighbouring cells. Leaving this issue for a later discussion, let us concentrate on the N channels available for a cell. Let us also assume that the traffic per subscriber is e Erlangs. Therefore the number of subscribers that can be served in the cell works out to N/e and Subscriber Density (SD) that can be served in this cell is approximately* ,

Thus, subscriber density is inversely proportional to the square of cell radius. The implication of this can be seen by a example: For e.g., if N = 200, e = 0.1 Erlang, the capacity (subscriber density) varies with cell radius as follows:

r = 25 km,

SD » 1 per sq. km

r = 10 km,

SD » 6 per sq. km

r = 3 km,

SD » 70 per sq. km

r = 1 km,

SD » 640 per sq. km

r = 500 m,

SD » 2550 per sq. km

Therefore, cell radius plays the dominant role in determining the subscriber density given a certain frequency spectrum. In other words, a smaller cell radius is the key to efficient use of spectrum and one may have to use cell radius as small as 500m, if one desires a reasonable subscriber density.

4.5 Choice of Multiple Access

A key parameter determining the efficient reuse of spectrum is governed by
multiple-access technique used. The access technique defines how the frequency spectrum is divided into channels and affects reuse of the channels.

4.5.1 FDMA

The oldest technique used in wireless access, especially in mobile communications, is Frequency Division Multiple Access (FDMA). Here the available frequency spectrum is divided in a number of orthogonal frequency channels and these channels are assigned to the user on demand. FDMA can be used both for analog as well as a digital communications. This simple technique used extensively in first generation analog mobile system, however, had poor reuse and the same channels can be reused only once in 14 or 21 cells. One way to increase re-use efficiency is by employing sectored or directional antennas at the cell site. (A brief discussion on sectorization will follow in section 4.6.1) Even with sectorisation, say 3 sectors per cell the best planning gives a typical reuse of once in 7 cells [4], implying reuse factor of 1/7 = 0.143 per cell.

4.5.2 TDMA

The most widely used multi-access technique today, both for mobile as well as in wireless local loop, is Time-Division Multiple Access (TDMA). Here the frequency spectrum available is again divided, but into a few (wide) bandwidth channels or carriers. Each carrier is used for transmission of multiple time-multiplexed channels. Each such orthogonal channel (or time-slot as is commonly referred to) could be assigned to a user on demand. The technique can be used only for digital communication, and the ability to work with smaller signal to interference ratio in digital domain, gives this technique better reuse factor as compared to the analog FDMA. For example, with three sectors, a cell reuse factor of 1/4 or even 1/3 is achievable [4].

4.5.3 CDMA

Late in the eighties emerged a multiple-access technique referred to as Direct Sequence, Code Division Multiple Access (DS-CDMA). Based on spread spectrum techniques used extensively in defense applications for over twenty years, this technique enables definition of near-orthogonal channels in code-space. CDMA enables multiple channels to use the same frequency and time slots. Each bit to be transmitted by or for a user is uniquely coded by spreading the bit into 64 or 256 or even 1024 chips. The receiver separates the data of a user by a decoder which correlates the receive signal with the code vector associated with that user. On correlation, the interference from other users would become nearly zero and add only a small amount of noise, where as the desired signal will be enhanced considerably. The technique is useful in exploiting the inherent time-diversity from multipath delay-spread, especially if the spreading is significant (Chip time of 0.1 m sec to 1 m sec. The only problem with the technique is that as completely orthogonal codes are not possible, especially on the uplink, the total bit-rate supportable from all users using this technique is significantly less than the total bit-rate supportable with TDMA and FDMA technique using the same frequency spectrum.

This disadvantage in the CDMA system is made up by better reuse efficiency, as the same spectrum with different set of codes can almost totally be reused in every cell. The theoretical reuse efficiency could be as high as 1.0, but in practice less. With sectored antennas, it is possible to reuse the spectrum in each sector, with a 3-sector cell site resulting in a reuse efficiency of nearly 0.5 per sector.

An issue that is as important as reuse of frequency spectrum is fine power control so that more or less equal power from each subscriber set reaches a base station. Such a control mechanism was difficult to implement and delayed widespread use or CDMA for sometime. Fortunately, the problem has been largely overcome today.

4.5.4 MC-TDMA

One of the latest access techniques that has emerged is the Multi-Carrier-Time Division Multiple Access or MC-TDMA with Dynamic Channel Selection (DCS). MC-TDMA is a variation of TDMA. A time frame is divided into time slots, as in TDMA; however, in each time slot, a subscriber equipment or Base Station can use any of the several frequencies available. Therefore as in TDMA, the spectrum available is divided into a set of frequencies. Each frequency or carrier can be used in anytime slot by communicating equipment. The key is that no frequency or time-slot is assigned to any subscriber equipment. Nearly all channels are available as a pool for every one to choose from. A technique known as Dynamic Channel Selection (discussed in more detail in section 4.5.5) governs the choice of the channel and is key to high reuse efficiency. A reuse efficiency of 0.7 to 0.8 may be possible from cell to cell and with a 3 sector cell, one may get a reuse efficiency of almost 0.7 per sector.

4.5.5 Fixed Channel Allocation (FCA) Versus Dynamic Channel Selection (DCS)

Most wireless access systems till recently used Fixed Channel Allocation which required a prior allocation or assignment of certain number of channels (carrier frequencies) to a sector in a cell using an exercise generally referred to as frequency planning. The planning had to be carried out using a worst case scenario assuming the nearest possible distance between the interfering signals. Having carried out this worst case planning, and having assigned the channel pool to the base station serving the sector, it was upto these base stations (or fixed part) to assign channels to subscribers in the sector on demand.

A totally different approach emerged recently [5], initially as a theoretical concept and as soon implemented in a number of systems. This approach, called Dynamic Channel Selection (DCS), does no assignment of channels to any base station or subscriber equipment. All channels are available to every one. The radio equipment is designed to measure the signal strength that it receives on all channels (using something akin to a spectrum analyser) and thus determine the actual radio environment in its vicinity. It carries out this measurement on a continuous basis, whether it is using a channel or not. Thus the complete knowledge of the radio environment enables it to select a channel in which it can communicate best. The key is that even while it is communicating on one channel, it is measuring the radio-environment in all other channels. If it finds that another channel can provide it better communication, it switches to this channel seamlessly.

Thus the radio equipment (usually associated with the subscriber equipment) continuously monitors all the channels, and selects the best channel dynamically. It tries to establish communication on this seemingly best channel. If it succeeds, so be it; otherwise it tries the next best channel.

The DCS is thus based not on worst case scenario, but on actual radio environment. It is thus possible sometimes to reuse a channel even 25 m from the other. One such case is shown in Fig.5 (a) and 5(b). In this case two base stations, B1 and B2, are located only
25 m apart. In scenario (a), the two subscriber equipments (handset or HS) referred to as HS1 and HS2, are located close to each other. The HS1 is communicating to BS1 and HS2 with BS2. The two communication can not reuse same channel and use different duplex channels C1 and C2. However, in scenario (b), the HS1 and HS2 can communicate on same duplex channel C1. This is because HS1 is about 2 m from BS1 but 25 m from BS2. The interference that it receives from BS2 (on the same channel) is approximately (25/2)2 or nearly 22 dB less than the signal it receives from BS1. This interference level causes no problems. Same is the situation for reception by HS2 from the two base stations.

This reuse even 25m apart is possible because of DCS. No FCA with worst case planning can come close to this in reuse. DCS gives a factor of 2 to 4 in reuse advantage compared to FCA [6].

It is the DCS which gives techniques like MC-TDMA an edge over other multiple-access systems that, makes its reuse efficiency on the average very high. Of course, DCS requires fairly sophisticated radio environment measurement techniques in each radio equipment including subscriber equipment. However, such sophistications can easily be included in today's integrated circuits and digital signal processors.

4.6 Interference Reduction Techniques

The re-use distance is primarily determined by the target Signal to Interference Ratio (SIR) requirement. The target SIR is based on the minimum sensitivity required at the receiver input in order to obtain a particular Bit Error Rate (BER). The required BER is typically 10-3 for voice applications (and 10-6 or higher for data applications by using error control coding and/or ARQ). Depending on the choice of multiple access, the modulation scheme and the particular application (mobile or fixed wireless), the target SIR set point will differ.

Interference reduction techniques are widely used in wireless systems to increase re-use efficiency while retaining the target SIR requirement [7]. These techniques include:

  1. Sectorization
  2. Voice Activity Detection
  3. Power Control
  4. Rate Control
  5. Frequency Hopping

4.6.1 Sectorisation

It is possible to sectorise a cell by using base stations with directional antennas, such that the base station serves subscribers only in that sector as shown in Fig. 6. Such
non-overlapping sectors can not only reduce the interference power, but also increase range. It is also possible to use overlapping sectors in order to improve trunking efficiency, and thereby, support more users in an area. Overlapping sectors are more suitable for fixed wireless applications.

Directional antennas providing 600 sectors have been used in GSM and IS-136 deployments to increase the re-use ratio from 1/7 to 1/4 and even 1/3 (ie., an increase from 7-cell reuse to 4cell and 3cell reuse). A capacity increase of nearly 4.5 times over omni cell capacity has been obtained using 600 sectoring in IS-95 deployments. The reuse is better if the antenna at the base station has high loss outside the sector(s) that it is supposed to serve. The subscriber end antenna can also be directional in case of fixed (non-mobile) subscriber installations. This helps immensely to reduce interference power especially when the subscriber is near the edge of the cell. Ofcourse, directional antennas are not preferable on the mobile site since they affect handoff performance.

      1. Other Interference Reduction Techniques

In continuously transmitting wireless systems like DS-CDMA, Voice Activity Detection (VAD) is very useful to reduce power consumption, and also increase user capacity. In VAD, not only is the presence or absence of speech energy monitored, but also the regions corresponding to unvoiced speech and the transition regions between voiced to/from unvoiced speech. IS-95 exploits VAD to reduce its bit-rate from 8kbps to 4/2/1kbps corresponding to unvoiced speech, transition regions, and comfort noise regions. While the 13kbps GSM codes also incorporates VAD, it is not explicitly exploited in reducing power or bit-rate.

VAD information can be used to define rate control and/or power control operations. For example, in IS-95, when VAD returns a no-speech activity flag, the bit-rate for that user can be dropped from 8kbps to 1kbps (rate control) by lengthening the bit period by a factor of eight (from Tb to 8Tb). Simultaneously, the transmit power of the user can be reduced by a factor of 82 = 64. Now, the integrator in the receive will integrate over 8Tb (and not Tb), which will give back the same energy per bit for that signal. Therefore, rate and power control do not charge the effective energy-per-bit or SNR for that signal.

Ofcourse, sophisticated power control is also done (both on the base and mobile stations) in IS-95 to mitigate the near-far problem of direct sequence spectrum communications. Even TDMA/FDMA systems can employ rather simple power-control mechanisms to reduce co-channel and adjacent channel interference. For example, all low range (near base station) duplex channels can lower their peak powers and still meet the target SIR in GSM systems.

Another way to reduce co-channel interference in FCA applications is to have a pre-set slow frequency hopping between the base stations in the region. Thus slow frequency hopping, typically once every second or so, is synchronised by the mobile switching centre which controls these base stations. The effort of this slow hopping of the frequency allocation is to randomize the geographical location of the co-channel signal(s), and thereby, on the average reduce interference and increase re-use efficiency.

4.7 Capacity

Thus, there are several factors, which determines the efficient use of a radio spectrum in a wireless in local loop system. In this section, we will attempt to determine an expression for the capacity, C, of a WiLL system. The capacity expression will be in terms of the number of subscribers per sqkm that can be served given an available spectrum, Erlang traffic used, cell radius and number of sectors used and the payload that has to be provided to each subscriber. Some of the popular access techniques will be compared using this capacity expression. Let us start with a few definitions:

Capacity = C = Numbers of subscribers served per sqkm

Multiaccess and modulation efficiency and overheads = M bps/hz: This factor combines modulation efficiency, effect of overhead and multi-access efficiency and gives the number of bits of payload delivered per Hz of spectrum.

r = radius in km for each cell

ns = number of non-overlapping sectors used per cell

R = reuse efficiency (depends on access technique and ns, and governs fraction of
total spectrum that can be effectively used in each cell and each sector)

e = Erlang traffic per subscriber

Te = trunking efficiency (a factor which depends on the Erlang traffic per
subscriber, e, and number of channels available in each sector/cell)

S = total spectrum 9available in Hz

x = payload in bps required per subscriber.

Therefore, total number channels of x bps available is total spectrum available multipled by modulation and multi-access efficiency and divided by payload required per subscriber or (SM/x).

Since R fraction of these channels can be utilised in each sector, the number x bps channels available per sector (SMR/x).

Since traffic per subscriber is e Erlangs, if the number of channels available in each sector is large, the total number of subscribers in each sector would have simply been the number of channels available divided by e. However, if the number of channels per sector is not large, the trunking efficiency, Te, will reduce the number of subscribers that can be served in each sector.

Therefore, number of subscribers in each sector

Since there are ns non-overlapping sectors per cell, and r is the radius of the cell, the capacity or subscriber density C that can be served is

subscribers/sq.km (2)

Note that not all variables on the right hand side of the equation are independent as R is a function of access technique and ns, whereas Te is a factor depending on (SMR/x) as well as to some extent on the non-ideality of the sectored antennas.

5 Cellular and Wireless Standards

Having discussed various parameters which affects the efficient use and reuse of frequency spectrum, let us take a brief look at some of cellular and wireless standards that have emerged in the last twenty years. In section 5, we will look at the capacity provided by some them.

5.1 AMPS and NMT

First generation cellular system, which promised wide-scale mobile communications emerged in early eighties in the form of NMT-450 (Europe, 1981), NTT (Japan, 1978), and AMPS (USA, 1983). These were analog FDMA systems and used 400 and/or 900 MHz frequency spectrum to provide analog voice connection to mobile users [7].

5.2 GSM and D-AMPS

In late 80's emerged the second generation mobile systems. These systems were digital and mostly used TDMA. GSM [8], [9] was the most prominent amongst these and used 13.6 kbps voice coding. Initially designed for 900 MHz operation, the systems are now available in 1800 MHz and 1900 MHz in the name of DCS1800 or DCS1900. The GSM system is by far the most dominant system used in the world today.

A second TDMA system that emerged in the USA at about the same time is the Digital AMPS or D-AMPS [10] and was standardised as IS-54. The system was designed to maintain compatibility with analog AMPS system and used three 8 kbps time slots on a 30 kHz band used in AMPS. In mid-nineties, IS-54 standard was further refined and generalised and is today known as ASI-136. This has a proposal to incorporate micro-cell architecture and Dynamic Channel Selection.

5.3 IS-95

In early 90's, a CDMA standard emerged for mobile communication [11] in the form of IS-95. The standard used a 8 kbps (and later, a 13 kbps) voice coder, and was initially designed to operate in 900 MHz band. The IS-95 mobile system is widely used in S.Korea and parts of N.America, where GSM system was initially not allowed to operate. The IS-95 system is now being proposed to be used as a Wireless Local Loop solution in some countries. A 1800 MHz version IS-95 has also been proposed [12].

5.4 DECT

In early 90's another wireless standard emerged in Europe, called DECT [13]. This MC-TDMA system, initially proposed for home cordless and office PBX market, was later adopted for wireless in local loop system. The system is defined for the 1800 MHz band, and employs 32 kbps ADPCM coding for voice and allows 64 kbps and even higher-rate data communication. The system uses Dynamic Channel Selection [14] to enhance frequency reuse. Similarly, the Micro-cellular system PHS [15] was standardised in Japan and PACS [16] was standardised in USA at about the same time. These systems were also proposed to provide communications with limited mobility in areas with very high teledensity.

5.5 Capacity provided by GSM, IS-95 and DECT system

In what follows, an attempt is made to determine capacity (as defined in section 4.7) of GSM, IS-95 and DECT systems. A comparison of the capacity is difficult because these systems were designed to provide communications in different environments. Even then, an attempt is made in this direction.

5.5.1 Modulation and multi-access efficiency

Lets begin with comparison of the factor M, the modulation and multi-access efficiency factor defined in section 4.7. Note that this also takes account the signalling overhead and determines the number of bps of payload per Hz of spectrum delivered by each technique.

GSM: Enables 8 channels each with 13 kbps payload using 200 kHz of spectrum. Obviously,

 

M32(GSM)  =  8 x 13 kbps  =  0.52bps/Hz
  2000kHz

 

IS-95: 25 voice channels of 8 kbps without Voice Activity Detection (no VAD) in 1.25 MHz spectrum [11]. Therefore,

 

M8,noVAD(IS-95) = 25 x 8 kbps = 0.16bps/Hz
     1250kHz

40 voice channels of 8 kbps with VAD and silence suppression in 1.25 MHz spectrum. Therefore,

 

 

M8,VAD(IS-95) =  45 x 8 kbps  = 0.256bps/Hz
      1250kHz

 

25 voice channels of 13 kbps with VAD and silence suppression in 1.25 MHz spectrum. Therefore,

 

 

M13,VAD(IS-95) =  25 x 13 kbps  = 0.26 bps/Hz
      1250 kHz

 

DECT: 120 channels of 32 kbps full duplex in 20 MHz band.

 

 

M32(DECT) = 120 x 32 x 2 kbps = 0.384bps/Hz
   20,000 kHz

 

It is obvious that modulation and multi-access efficiency is higher for GSM than for DECT or IS-95. But IS-95 and DECT perform better as compared to GSM when it comes to reuse efficiency, as discussed in section 5.5.2. DECT uses less efficient modulation (GMSK with BT = 0.5) which can be implemented at a very low cost. It also uses a large signalling overhead. However, the dynamic channel selection in DECT, more than compensates for these deficiencies.

5.5.2 Subscriber Densities for Various Systems

Let us now look at the capacity of each system. For comparison purposes, let us assume 20 MHz spectrum (10 MHz for uplink and 10 MHz for downlink) is available for each system. Let us take Erlang traffic per subscriber to 0.15 E/subscriber, taking into account that WiLL today is not only used for telephone, but also for Internet service. What does each type of system enable us to do?

5.5.2.1 GSM Capacity

GSM uses 13 kbps voice communication. In 20MHz of total spectrum (paired spectrum of 10MHz eachway), 400 channels are available. When using non-sectorised cells, at best a reuse efficiency of 0.33 is possible. This implies, about 400 x 0.33 or 135 channels per cell implying a trunking efficiency of about 0.85. Thus the number of subscribers per cell works out to be approximately 766. For a cell radius of 10 kms, 3 kms and 1 km, of the subscriber density served given by 766/(p r2) will be 2.4, 27, and 245 subscribers/sqkm, respectively.

Using three sector deployment with 120o sectors, the reuse efficiency will be closer to 0.2 per sector per cell. This would give 400 x 0.2 or 80 channels per sector giving a trunking efficiency of 0.8. This works out to be 80/0.15 or 425 subscribers in each sector of a cell. Therefore for cell radius of 10 kms, 3 kms and 1 kms, the subscriber density supported is 425/(p r2) or 4, 45, and 410 subscribers/sqkm, respectively.

5.5.2.2 IS-95 Capacity

IS-95 uses either 13 kbps or 8 kbps voice communications with or without VAD. For use of 8 kbps with VAD, 20 MHz spectrum (for two way communications) would support 10,000 x 0.25/8 or 320 voice communications (using the fast that M8, VAD (IS-95) = 0.256 bps/Hz) simultaneously. For a non-sectorised cell, with reuse efficiency R = 0.7, nearly 225 channels are available. The trunking efficiency for these many channels would be 0.9 implying number of subscribers per cell would be 225 x 0.4/0.15 or 1350 and for cell radius of 10 km, 3 km and 1 km* * , subscriber density would be 4.3, 47 and 429 subscribers/sqkm respectively. For 3 sector cell, reuse efficiency is approximately 0.5 for each sector. The trunking efficiency would again be 0.9 and the subscriber density served would approximately be 9.2, 102, and 920 subscriber/sqkm, respectively.

For 8 kbps communication without VAD, 20 MHz spectrum would support 200 voice communications simultaneously. The trunking efficiency would now reduce to 0.85 and 0.82 for non sectorised and 3 sector deployment and the subscriber density supportable for 10 km, 3 km and 1 km would be 2.5, 28, 252 subscribers per sqkm respectively for non-sectorised deployment and 5.2, 58, and 522 subscribers/sqkm for 3 sector deployment.

For 13 kbps communication with VAD, 20 MHz spectrum would again support 200 voice channels and subscriber density supported would be approximately same as that in 8 kbps without VAD case.

5.5.2.3 DECT capacity

DECT supports 120 full duplex channels of 32 kbps in 20 MHz spectrum. The reuse factor is about 0.7 for non-sectorised deployment, implying 85 channels in a cell.

For 85 channels, the trunking efficiency should have been 0.8. However, due to dynamic channel selection and due to all 120 channels being available to every sector/cell, an average trunking efficiency of 0.9 is more likely to be achieved. The number of subscribers per cell would therefore be 85 x 0.9/0.15 or 510, and the subscriber density supportable for 10 km, 3 km, 1 km and 500 m radius would be 1.6, 18, 162 and 650 subscribers/sqkm. Cell radius of 500 m has been taken as this is possible in DECT, whereas it is very difficult and expensive to implement it in GSM or IS-95.

For a three sector deployment, reuse efficiency is 0.5 for each sector implying 60 channels in a cell. Again, trunking efficiency would remain close to 0.9. The subscriber density supportable would therefore be 3.4, 38, 343 and 1375 subscribers per sqkm. Tables 5.1 and 5.2 tabulates these results for GSM, IS-95 and DECT.

 

9Cell-radius km10km

2.4

2.5

4.3

1.6

3km

27

28

47

18

1km

245

252

429

162

500 m

--

--

--

650

Table 5.1: Subscriber density (subscriber/sqkm) supportable in single sector deployable
spectrum = 20 MHz, Erlang traffic = 0.15 E/subscriber

Cell-radius km

GSM (13)

IS-95(8, noVAD) and
IS-95 (13, VAD)

IS-95
(8, VAD)

DECT (32)

10km

4

5.2

9.2

3.4

3 km

45

58

102

38

1 km

410

522

420

343

500 m

--

--

--

1375

Table 5.2: Subscriber density (subscriber/sqkm) supportable in 3-sector deployment
(Spectrum = 20 MHz
, Erlang traffic = 0.15E/subscriber)

From the results in Table 5.1 and Table 5.2, one can easily conclude

    1. microcell is a must to support high subscriber density. Only cells with 1 km or less can give the type of subscriber densities required in Indian urban areas.
    2. DECT with 32 kbps payload supports only slightly lower subscriber density as compared IS-95 with 13 kbps (with VAD). This implies DECT has higher spectral efficiency.

5.5.3 Capacity Comparisons for Providing 64 kbps Data Connections

To compare apples with apples, let us take a look at a situation, where both DECT and IS-95 is used to provide 64 kbps full duplex data communication using 20 MHz BW. Without VAD, IS-95 would support only about 3 channels of 64 kbps in 1.25 MHz (each way) spectrum, implying 25 (64 kbps) channels for 20 MHz bandwidth. The trunking efficiency would now be approximately 0.6. Thus only 15 channels would be available per sector or cell, even assuming 100% reuse. On the other hand, DECT supports 60 channels of 64 kbps. The trunking efficiency would be close to 0.8 implying that 48 channels would be available per sector of a cell assuming once again 100% reuse.

6 The Future

Having looked at the fundamental issues of importance for WiLL systems and a brief look at available technological options, an obvious question that emerges is what lies in the future? We would like to conclude the article with a brief pointer towards some interesting on-going development. The first is in terms of incorporating space time processing and second in the emergence of 2.5G and 3G wireless technologies and its impact on future WiLL systems.

    1. Space-Time processing

The last decade has seen the emergence of many theoretical and practical techniques, which exploit the spatial dimension in a more effective way than more diversity combining [17]. One may classify these new techniques broadly under four heads, namely:
(i) Smart Antenna Technology, (ii) Transmit Diversity Schemes, (iii) Spatial Multiplexing and (iv) Space-Time Coding.

6.1.1 Smart Antennas

Smart antennas enable focus of beams from base station towards the subscriber terminal and vice-versa. Such electronic focussing using adaptive, phased-array antennas not only enables improvement in link margins, thus enabling longer reach with same power, but also has the capability of significantly improving spectrum reuse. The latter is because, two focussed pencil beams may use the same channel and still could produce very little interference to each other. Key is to have smart antennas provide such pencil beams for each user. The concept is extremely interesting and it has the potential of significantly improving spectral efficiency especially in fixed wireless applications. However, the work in this area is still in early stages with some commercial products incorporating some very early versions of smart antennas. It will take a few years for products providing significant space diversity to emerge.

6.1.2 Transmit Diversity

A signal is transmitted simultaneously from two antennas at the base station, on the same carrier. By appropriate coding of the data streams put on the two antennas, the receiver can reduce the fade margin significantly. It requires feedback from the mobile/portable to the base station to obtain the channel state information and optimize performance.

6.1.3 Spatial Multiplexing

Use of multiple antennas at a (fixed) subscriber terminal as well as at the base station allows one to spatially multiplex different data streams on the same carrier by transmitting the streams through different antennas. Use of L antennas can give (nearly) an L-fold increase in data rate. This technique is also called MIMO (multi-input multi-output) processing, one popular version of spatial multiplexing, called V-Blast [18] has been shown to yield spectral efficiencies as high as 30 bits/sec/Hz in short-range, fixed wireless application.

6.1.4 Space-Time Coding

In conjunction with MIMO processing, error-control coding can further increase range and/or bit-rate. This coding across spatial channels use ideas similar to Trellis Coded Modulation (TCM) to define mappings of symbols to antenna ports [19]. Some of these techniques are also being currently used to provide broadband wireless access to homes at rates upto 2Mbps.

6.2 3G Terrestrial Wireless Standards

The third-generation (3G) standardardisation activity was started by various groups after ITU announced the release of new spectrum for the International Mobile Telephone (IMT) application, and invited proposals for the same. This spectrum, in the 1.9-2.1 GHz band, could be utilised only by those standards which satisfy the IMT-2000 requirements. (see for example, [20], [21], and [22], for good overview articles on IMT-2000 and 3G technologies).

For the IMT-2000 terrestrial radio transmission technology, the main specification was that any compliant standard should provide data rates of 144kbps for mobile applications, 384 kbps data for pedestrian applications, and upto 2Mbps speed for fixed applications. In addition, all 3G systems are likely to support Generalised Packet Radio Service (GPRS), and interface with the GSM core network. Greater capacity and higher spectral efficiencies, support for multimedia services, incorporation of 2G services, interconnection with mobile satcom, and international roaming are a few other important requirements of IMT-2000.

By end 1998, about 11 proposals were submitted to ITU, of which 9 of them were direct sequence CDMA based and the other two were TDMA based (one which was a DECT evolution, and the other was a GSM evolution). However, by end 1999, many of these proposals were either withdrawn or merged to form 4 proposals, namely: (i) 3G Partnership Program (3GPP) Wideband CDMA (W-CDMA) proposal from Europe and Japan, (ii) Enhanced Data Rates for Global Evolution (EDGE) from USA and Europe, and (iii) Multicarrier CDMA or CDMA2000 from USA, (iv) EP-DECT from Europe.

All the 3G standards incorporate features to support higher peak data rates, multiple rate services, and multimode capabilities. For example, the 3G radio will be a coherent (or differentially coherent) one. It will support the new W-CDMA (3GPP) and/or CDMA2000 standard, as 9well as a fallback to either GSM or IS-136/IS-95. Support of EDGE is also a possibility, using multiband functionality. The W-CDMA standard has a channel spacing of 5 MHz and a chip rate of 3.84Mcps, while CDMA2000 has a chiprate in multiples of 1.2288Mcps is bandwidth multiples of 1.25MHz. EDGE, on the other hand has exactly the same slot and frame structure as GSM, but by changing the modulation from GMSK to 8-PSK, will have a gross bit-rate of 812.49 Kbps.

While W-CDMA will use the idea of Orthogonal Variable Spreading Factor (OVSF) codes to provide multiple rate services, EDGE would use a link adaptation procedure to provide variable bit rates. It is expected that the 3G systems will provide about twice the capacity of the current 2G systems, where most of the gain accrues from more efficient modulation and better int9erference reduction approaches. In addition, the 3G standard provides the requisite support for a number of new techniques that were invented in the last decade.

These include:

  1. Turbo codes
  2. Multi-user detection
  3. Space-time processing

6.2.1 Brief Background on the New Techniques

We have already discussed space-time processing in section 6.1, and therefore we will confine here our description to the other two techniques.

Turbo Codes: These new error correcting codes are compute intensive at the receiver. The idea of parallel concatenation of simple codes is used to define powerful error correction mechanisms. The performance improves as one repeatedly iterates the decoder. In return, one can reduce the SNR by 4-6 dB (even greater in many cases). Turbo coding is good for data, but the interleaver/deinterleaver and decoder delay is currently considered to be too high for voice applications.

Multi-User Detection: This is theoretically superior to co-channel suppression or cancellation techniques, and can increase capacity/re-use considerably. Multiuser detection techniques have been well studied for DS-CDMA signals, but relatively less known for TDMA signals. The pilot signals on the W-CDMA uplink enable fairly sophisticated multiuser detection at the base-station.

For the mobile cellular application, interoperability will be key, and multimode and/or multiband functionality with some of the above advanced techniques need to employed. To make these ideas a reality, a combination of powerful DSPs and ASIC accelerators are needed in the baseband section. The algorithms have to be crafted carefully, with an optimised hardware/software partition to minimise cost and power consumption.

On the other hand, it seems that for fixed wireless access, like in the wireless in local loop application, not much re-engineering of the 2G fixed wireless systems may be required in order to support IMT-2000 bit rates for pedestrian and fixed applications. As an example, we consider the evolution of DECT to 3G in the next section.

6.2.2 DECT evolution towards 3G

The 3G requirements are 384 kbps data for pedestrian applications, and 2Mbps speed for fixed applications. The DECT standard has been expanded to provide for these - only the 3G mobile requirements (of 144 kbps) are not met.

In particular, the extended DECT standard is very attractive for fixed wireless applications. In extended DECT, the gross bit-rate is tripled from 1.152Mbps to 3.456Mbps by replacing GMSK with 8-PSK modulation. Exploiting the asymmetry allowed by TDD systems, is quite straight forward to obtain 2Mbps shared downlink services on such a standard.

6.3 Internet Access using Wireless Networks

Fixed and mobile wireless network standards have been evolving to support increasingly higher data rates. Whereas the 2G systems provided for low bit-rate (9.6 kbps in the case of GSM) circuit-switched data connection, the interim 2.5G mobile system using GPRS and GSM enhanced the bit-rate to a reasonable high value of 115.2 kbps, by transmitting in multiple time slots and using circuit switched connection. Similarly the DECT-DPRS supports a much higher data rates than that of 24 kbps rate supported in DECT. Both GPRS and DPRS systems are inherently circuit switched connections and hence can only support a small number of data subscribers.

The 3G standards aim to support high datarates and a large number of active data subscribers. The EDGE, WCDMA and EP-DECT are the examples of 3G systems. The main features of these systems are:

  1. Large signal constellations: Instead of binary or QPSK modulation in 2/2.5G systems, these systems use 8-PSK modulation and thereby affect a three-fold increase in the raw bit rate.
  2. Decoupling of uplink and downlink streams: This enables one to communicate at much higher rates on the downlink compared to the uplink rate.

Asymmetry between uplink and downlink rates is a important feature of internet traffic.

  1. Packet Switched Connections: This exploits the inherent "burstiness" of the data traffic to support a large number of active users.
  2. Contention based access: The active users contend using random access protocols similar to ALOHA, to transmit on the uplink.
  3. Point-to-multipoint downlink: This feature allows for the downlink channel to be shared among all or a group of active users.

The data rates the 3G system support vary from a low of 8 to 32 kbps to a high of 1 to
2 Mbps. With contention based uplink schemes (in the case of DECT with half-slot connections also) it is possible to support large number of users to be connected.

7 Summary and Conclusion

In this paper, an attempt has been made to compare the subscriber densities that can be supported by major wireless access standards including GSM, IS-95 and DECT. The emphasis was on their applicability for the local loop, which implies not only the need to support very higher user densities, but also provide toll quality voice and simultaneously cater to high speed Internet requirements. In this context, it emerged from this study that for local loop applications, micro-cellular networks which can very efficiently re-use the frequency spectrum are absolutely essential in order to meet bit-rate demands of 64 kbps or higher and user densities greater than 1000 subscribers/sqkm. Micro-cellular standards like DECT or PACS use dynamic channel selection to allocate frequencies across base stations, which results in very efficient frequency reuse. This also directly brings down costs, but also allows for a flexible and easy expansion of the base station coverage regions.

Finally, this article also discussed some of the key ideas in the emerging third generation (3G) wireless standards. A brief description of the new technologies that 3G will introduce into the network, and what will be their impact on Internet and multimedia bit-rates and services, were also provided. Some important ideas that the air-interface would need to employ in order to cater to Internet traffic have been described

It is expected that with the advent of 3G wireless standards, not only will greater bit-rates be delivered to mobile users, but there will be a substantial improvement in the bit-rates and services for fixed wireless applications like the local loop and broadband to home as well.

 

References

  1. ITU-T Recommendation G.703, "Physical/Electrical characteristics of hierarchical digital Interface, 1996
  2. ITU-T Recommendations V5.2
  3. Ashok Jhunjhunwala, Bhaskar Ramamurthi, Timothy A. Gonsalves, "The Role of Technology in Telecom Expansion in India", IEEE Communication Magazine, November, 1998
  4. W.C.Y.Lee, "Mobile Communications Engineering", 2nd Edition Mc Graw Hill, New York 1998
  5. Reference in section 4.95 on DCS
  6. Cox D.C., "Wireless Network Access for Personal Communications", IEEE Communications Magazine, pp.96-114, December 1992.
  7. Rappaport, T.S., "Wireless Communications — Principles and Practice", Prentice Hall, New Jessey, 1996.
  8. Mouly, M., and Pautet, M.B., The GSM System for Mobile Comunication, ISBN: 2-9507190-0-7, 1992.
  9. V.K. Garg and J.E. Wilkes, "Principles and Applications of GSM", Prentice Hall, New Jessey,1999.
  10. EIA/TIA Interim Standard, "Cellular System Dual Mode Mobile Station — Land Station Compatibility Specifications", IS-54, Electronic Industries Association, May 1990.
  11. M. Gilhousen, et.al., "On the Capacity of Cellular CDMA System", IEEE Transactions on Vehicular Technology, Vol.40, No.2, pp.303-311, May 1991.
  12. ANSI J-STD-008 — Personal Station-Base Compatibility Requirements for 1.8-2.0 GHz Code Division Multiple Access (CDMA) Personal Communications Systems, March 1995.
  13. H. Ochsner, "DECT — Digital European Cordless Telecommunications", IEEE Vehicular Technology 39th Conference, pp 718 – 721, 1989.
  14. D.C.Cox, "Wireless Network Access for Personal Communications", IEEE Communications Magazine, pp. 96-114, December 1992.
  15. K. Ogawa, et. al., "Toward the Personal Communication Era — the Radio Access Concept from Japan", International Journal on Wireless Information Networks, Vol.1, No.1, pp.17-27, January 1994.
  16. D.C.Cox, W.Arnold, and P.T.Porter, "Universal Digital Portable Communications: A System Perspective", IEEE Journal on Selected Areas of Communications, Vol.SAC-5, No.5, pp.764, 1987.
  17. A. Paulraj and C.B. Papadias "Space-time processing for Wireless Communications", IEEE Signal Processing Magazine, vol.14, pp.49-83, Nov 1997.
  18. G.J. Foschini and M.J. Gans "On Limits of Wireless Communications in a fading Environment using MULTIPLE Antennas", Wireless Personal Communications, vol.6, No.3, March 1998.
  19. V. Tarokh, N. Seshadri and A.R. Calderbank, "Space-Time Codes for High Data Rate Wireless Communication: Performance Criterias and Code Construction", IEEE Transaction an Information theory, vol.44, No.2, March 1998.
  20. Special Issue on "Third Generation Mobile Systems in Europe", IEEE Personal Communications, vol.5, No.2, April 1998.
  21. Special issue on, "IMT-2000: Standards Efforts of the ITU", IEEE Personal Communications vol.4, No.4, August 1997.
  22. Special Issue on "UMTS" IEEE Transaction on Vehicular Technology, vol.47, No.4, November 1998.